DLL1-mediated ferroptosis resistance via the Notch-Nrf2/GPX4 axis drives cisplatin resistance in ovarian cancer Authors: Dan Ye, Gaoting Huang, Xingcheng Yan, Lezi Suntan, Haoran Shen, Jian Shen Published: 2026 ABSTRACT BackgroundOvarian cancer (OV) is a leading cause of cancer-related mortality, with cisplatin resistance being a major clinical challenge. This study investigates the role of the Notch ligand DLL1 in mediating ferroptosis resistance and its impact on cisplatin sensitivity in OV.MethodsMulti-omics data, clinical samples, and OV cell lines were used to assess DLL1 expression, its link to prognosis, and its effect on the Nrf2/GPX4 axis. CCK-8, clone formation, and ferroptosis assays evaluated the impact of DLL1 on cell behavior and cisplatin sensitivity.ResultsDLL1 was upregulated in OV tissues and correlated with poor prognosis. Its knockdown inhibited cell proliferation, migration, and EMT, while inducing ferroptosis, evidenced by increased lipid peroxidation and mitochondrial dysfunction. Mechanistically, DLL1 activated the Nrf2/GPX4 antioxidant axis. DLL1 depletion sensitized OV cells to cisplatin. In vivo, combining DLL1-targeted therapy with a ferroptosis inducer significantly reduced tumor growth in cisplatin-resistant models.ConclusionDLL1 drives cisplatin resistance in OV by enhancing ferroptosis resistance via the Notch-Nrf2/GPX4 axis. Targeting DLL1 alongside ferroptosis induction represents a promising therapeutic strategy, positioning DLL1 as a potential biomarker and target in drug-resistant OV.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40170-026-00444-3. Introduction Ovarian cancer (OV) is one of the deadliest gynecologic malignancies and its high mortality rate is mainly due to late diagnosis [1, 2], high recurrence rate and cisplatin resistance [3, 4]. As the first-line chemotherapeutic agent for OV, cisplatin exerts initial efficacy but its long-term clinical benefit is severely limited by the development of resistance [5, 6]. Studies have shown that the mechanism of drug resistance not only involves drug efflux pumps or enhanced DNA repair [7], but is also closely related to cellular metabolic remodeling and imbalance of redox homeostasis [8, 9]. Ferroptosis, a novel mode of programmed cell death characterized by lipid peroxidation accumulation and iron-dependent ROS bursts [10, 11], has recently been shown to be significantly associated with cisplatin resistance [12, 13]. Delta-like ligand 1 (DLL1) is a canonical ligand of the Notch pathway [14, 15]. In this study, we reveal that DLL1 drives cisplatin resistance in OV by mediating ferroptosis resistance via regulating the Nrf2/GPX4 axis, providing new insights into chemoresistance mechanisms and a theoretical basis for ferroptosis-targeted combination therapy. Approximately 70% of OV patients experience recurrence after initial treatment, with platinum resistance being a key factor in treatment failure [16, 17]. Clinical data show that the 5-year survival rate of cisplatin-resistant patients is much lower than that of sensitive patients (approximately 50%) [18, 19]. In recent years, the regulatory mechanism of ferroptosis has become increasingly important in tumor research, and its core molecule GPX4 [20], maintains cell survival by reducing lipid peroxides [21], while GPX4 inhibitors (e.g., RSL3) have been shown to reverse a variety of tumor drug-resistant phenotypes [22–24]. Although the DDL1-Notch axis has been implicated in cisplatin resistance [25], it remains unclear whether this signaling axis influences cisplatin sensitivity in OV by regulating ferroptosis. The key scientific questions addressed in this study include: the expression characteristics and clinical significance of DLL1 in cisplatin resistance OV; how DLL1 regulates the key pathway of ferroptosis through the downstream target of Notch signaling; and whether targeting DLL1 can sensitize chemotherapy by inducing ferroptosis. Addressing these questions is critical to overcoming the existing bottleneck of drug-resistant OV therapy. Recent studies have gradually revealed the dual role of Notch signaling in tumor drug resistance: activation of the Notch1 receptor promotes tumor stemness [26, 27], while the VEGFA/Dll4/Notch1 pathway shows ability to inhibit tumor angiogenesis [28]. In contrast, the field of ferroptosis focuses on the xCT-GSH-GPX4 axis [29, 30] and ACSL4-mediated reprogramming of lipid metabolism [31], but the upstream regulators of the pathway are unclear. In particular, several studies have suggested that the Notch pathway may regulate oxidative stress through Nrf2 [32, 33], but whether DLL1 as a Notch ligand is involved in this process has not been reported. Furthermore, although GPX4 has been recognized as a core ferroptosis molecule, its dynamic regulatory mechanism in cisplatin resistance in OV remains to be elucidated. This study provides the first functional link between the DLL1-Notch-Nrf2/GPX4 signaling axis and ferroptosis resistance (Fig. 1). We demonstrate that DLL1 activates Notch signaling in a spatial-specific ligand-receptor manner to promote Nrf2-dependent GPX4 expression, and inhibits ferroptosis by preserving mitochondrial integrity and iron metabolism homeostasis. Targeting DLL1 combined with ferroptosis inducers significantly reverses cisplatin resistance. Different from previous studies focusing on Notch receptors or downstream effectors, our study explores chemoresistance from the ligand perspective, offering a target for precise intervention. We propose a “spatial signaling-metabolic remodeling” model: in cisplatin-resistant OV cells, DLL1 activates Notch receptors on adjacent tumor cell membranes in an autocrine/paracrine mode, triggering the release of γ-secretase-dependent intracellular structural domains (NICDs) [34], which bind to the Nrf2 promoter and promote its transcriptional activity after nuclear translocation. The GSH/GPX4 axis establishes an antioxidant barrier by scavenging lipid ROS and antagonizing cisplatin-induced ferroptosis. Notably, DLL1 simultaneously inhibited Fenton reaction-driven lipid peroxidation to enhance iron storage and reduce free Fe²⁺ accumulation. This model integrates the spatial regulatory properties of Notch signaling with the kinetic balance of iron metabolism, providing a novel explanation for the mechanism of drug resistance. Fig. 1DLL1 activates the Nrf2/GPX4 axis through the NOTCH signaling pathway, leading to ferroptosis resistance and cisplatin resistance in OV cells. In this study, we propose a novel molecular regulatory model: in cisplatin-resistant OV cells, the transmembrane ligand DLL1 triggers the classical Notch signaling cascade by forming a spatially specific ligand-receptor pair with the Notch receptor on the adjacent cell membrane through an autocrine/paracrine mechanism. The core mechanism consists of three key levels: first, protein hydrolytic processing and nuclear translocation of the Notch intracellular structural domain activates the transcriptional activity of Nrf2; second, Nrf2 promotes the expression of GPX4 by facilitating the synthesis of GSH; and ultimately, the GSH/GPX4 axis builds up a strong antioxidant barrier by scavenging lipid ROS to antagonize ferroptosis signaling pathway activation. Notably, although cisplatin reduced GSH levels by blocking glutathione reductase function, in drug-resistant cells with over-activation of the DLL1-Notch-Nrf2 axis, a dynamically balanced redox microenvironment was formed due to enhanced compensatory GSH synthesis and GPX4 overexpression, which resulted in a significant attenuation of chemotherapeutic drug-induced ferroptosis effects. This model innovatively elucidates that DLL1 reprograms cellular redox homeostasis through spatial signaling regulation, and then constructs a new mechanism of cisplatin resistance in OV, which provides a theoretical basis for targeting ferroptosis to overcome chemotherapeutic resistance DLL1 activates the Nrf2/GPX4 axis through the NOTCH signaling pathway, leading to ferroptosis resistance and cisplatin resistance in OV cells. In this study, we propose a novel molecular regulatory model: in cisplatin-resistant OV cells, the transmembrane ligand DLL1 triggers the classical Notch signaling cascade by forming a spatially specific ligand-receptor pair with the Notch receptor on the adjacent cell membrane through an autocrine/paracrine mechanism. The core mechanism consists of three key levels: first, protein hydrolytic processing and nuclear translocation of the Notch intracellular structural domain activates the transcriptional activity of Nrf2; second, Nrf2 promotes the expression of GPX4 by facilitating the synthesis of GSH; and ultimately, the GSH/GPX4 axis builds up a strong antioxidant barrier by scavenging lipid ROS to antagonize ferroptosis signaling pathway activation. Notably, although cisplatin reduced GSH levels by blocking glutathione reductase function, in drug-resistant cells with over-activation of the DLL1-Notch-Nrf2 axis, a dynamically balanced redox microenvironment was formed due to enhanced compensatory GSH synthesis and GPX4 overexpression, which resulted in a significant attenuation of chemotherapeutic drug-induced ferroptosis effects. This model innovatively elucidates that DLL1 reprograms cellular redox homeostasis through spatial signaling regulation, and then constructs a new mechanism of cisplatin resistance in OV, which provides a theoretical basis for targeting ferroptosis to overcome chemotherapeutic resistance These findings highlight the central role of DLL1 in chemoresistance and propose a new ferroptosis-sensitizing combination therapy paradigm. Our study expands the understanding of cisplatin resistance from “mutation-driven” to “redox homeostasis remodeling”, providing a new framework for precision OV therapy. Future studies will explore the epigenetic mechanism of DLL1-Nrf2 interaction and nano-delivery system-based DLL1-targeted therapy, which may accelerate the clinical translation of chemoresistance treatment for OV via interdisciplinary integration. Results Result 1: DLL1 is significantly elevated in OV and is associated with malignant disease progression In order to systematically analyze the biological role of DLL1 in OV, this study explored its expression pattern and functional relevance in depth through a multi-dimensional experimental system. Firstly, transcriptome data analysis based on the GEO database revealed that in the GSE6008 dataset, the expression level of DLL1 was significantly up-regulated in tumor tissues compared with normal ovarian tissues (p = 0.00075, Fig. 2A). This finding corroborated with the analysis of the GSE148003 dataset: in cisplatin-resistant OV samples, DLL1 expression was also elevated in cisplatin-resistant OV samples compared with normal tissues (p = 0.026), suggesting that this molecule may be potentially relevant to the clinical dilemma of chemotherapy resistance (Fig. 2A). Of interest, dynamic monitoring of the neoadjuvant chemotherapy cohort (GSE201600) showed that after receiving carboplatin/paclitaxel combination chemotherapy, the expression level of DLL1 in patients’ tumor tissues was slightly elevated compared with that before treatment (Figure S1A), a regulatory phenomenon suggesting that DLL1 expression may be directly involved in the mechanism of chemo-sensitivity regulation. Survival analysis further revealed the clinical prognostic value of DLL1. Kaplan-Meier curves showed that patients with high DLL1 expression (DLL1high) had significantly poorer clinical outcomes compared to those with low DLL1 expression (DLL1low). The overall survival (OS) of OV patients in the DLL1high group was significantly worse compared to those in the DLL1low group (p = 0.0165, Hazard Ratio = 1.779, 95% CI: 1.093–2.89, Fig. 2B) based on TCGA data. Similarly, the Disease-Free Survival (DFS) of patients with high DLL1 expression was significantly lower than that of patients in the DLL1low group (p = 0.0258, Hazard Ratio = 1.710, 95% CI: 1.052–2.77, Figure S1B), suggesting the potential of the clinical application of DLL1 in the assessment of OV prognosis. Fig. 2Multi-omic profiling reveals elevated DLL1 expression correlates with poor prognosis and cisplatin resistance in OV. A. DLL1 expression in normal and tumor tissues of OV according to GEO dataset GSE6008; DLL1 expression in OV patients from OV patients receiving platinum-paclitaxel neoadjuvant therapy who developed cisplatin-resistant or non-resistant individuals according to GSE148003 dataset; B. Kaplan-Meier survival curve showing overall survival (OS) in OV patients with high versus low DLL1 expression. Patients with high DLL1 expression (DLL1high) exhibit significantly worse OS compared to those with low DLL1 expression (DLL1low) (p = 0.0165, Hazard Ratio = 1.779, 95% CI: 1.093–2.89) based on TCGA data; C. Ovarian epithelial cells IOSE80 were used as a control, and western blot was used to detect the protein expression level of DLL1 in OV cell lines HEY, A2780, and SK-OV-3; D. OVs were collected in three groups of independent cases (Case1-3) tumor tissues (T) and normal tissues (N) adjacent to the cancer, and western blot was performed to detect the protein level of DLL1 over in the clinical samples; E. IHC staining using DLL1 antibody was performed to detect the positive expression and distribution of DLL1 in the clinical samples; F. Volcano plot of differentially expressed genes between DLL1-high and DLL1-low TCGA-OV tumor samples, p-value = 0.05, log fold change ≥ ± 0.5; G. KEGG pathway enrichment analysis based on the DLL1-associated differentially expressed genes from the TCGA-OV cohort. Corresponding clinical sample details are summarized in Table S1 Multi-omic profiling reveals elevated DLL1 expression correlates with poor prognosis and cisplatin resistance in OV. A. DLL1 expression in normal and tumor tissues of OV according to GEO dataset GSE6008; DLL1 expression in OV patients from OV patients receiving platinum-paclitaxel neoadjuvant therapy who developed cisplatin-resistant or non-resistant individuals according to GSE148003 dataset; B. Kaplan-Meier survival curve showing overall survival (OS) in OV patients with high versus low DLL1 expression. Patients with high DLL1 expression (DLL1high) exhibit significantly worse OS compared to those with low DLL1 expression (DLL1low) (p = 0.0165, Hazard Ratio = 1.779, 95% CI: 1.093–2.89) based on TCGA data; C. Ovarian epithelial cells IOSE80 were used as a control, and western blot was used to detect the protein expression level of DLL1 in OV cell lines HEY, A2780, and SK-OV-3; D. OVs were collected in three groups of independent cases (Case1-3) tumor tissues (T) and normal tissues (N) adjacent to the cancer, and western blot was performed to detect the protein level of DLL1 over in the clinical samples; E. IHC staining using DLL1 antibody was performed to detect the positive expression and distribution of DLL1 in the clinical samples; F. Volcano plot of differentially expressed genes between DLL1-high and DLL1-low TCGA-OV tumor samples, p-value = 0.05, log fold change ≥ ± 0.5; G. KEGG pathway enrichment analysis based on the DLL1-associated differentially expressed genes from the TCGA-OV cohort. Corresponding clinical sample details are summarized in Table S1 To validate the results of bioinformatics analysis, systematic experimental validation was carried out. Western blot assay showed that DLL1 protein showed specific high expression in OV cell lines including A2780 and SK-OV-3 (Fig. 2C), with the expression of SK-OV-3 cells reaching 1.8-fold of that in normal ovarian epithelial cells (p < 0.001). At the level of clinical samples, we found that DLL1 protein expression was significantly higher in tumor tissues (T) than in normal tissues (N) in all three independent groups of cases (Case1-3) by paired analysis (Fig. 2D-E). Immunohistochemical quantification showed that DLL1 positivity was significantly higher in tumor regions than while normal tissues (Table 1). This multi-level expression consistency strongly confirmed the key role of DLL1 in OV development. Table 1Relationship between DLL1 expression and tumor characteristics in patients with OVFeaturesNo. of patientsDLL1 expressionp valueLowHighAll patients1267056Age (years)0.7620≤ 50643529>50613526Grade< 0.0001***I & II28244III743044stage0.19231 & 23724133 & 4904743Tumor size0.9491<12.9 cm643331≥ 12.9 cm713734T Infiltrate0.2338T1 & T2362313T3904743Lymphatic metastasis(N)0.0788N0955738N1311318Metastasis0.1940M01015942M1251114Recurrence of state0.3426No25169Yes1015447According to Mann-Whitney U analysis, there was a significant difference between the expression of DLL1 gene and the pathological grade Table 1Relationship between DLL1 expression and tumor characteristics in patients with OVFeaturesNo. of patientsDLL1 expressionp valueLowHighAll patients1267056Age (years)0.7620≤ 50643529>50613526Grade< 0.0001***I & II28244III743044stage0.19231 & 23724133 & 4904743Tumor size0.9491<12.9 cm643331≥ 12.9 cm713734T Infiltrate0.2338T1 & T2362313T3904743Lymphatic metastasis(N)0.0788N0955738N1311318Metastasis0.1940M01015942M1251114Recurrence of state0.3426No25169Yes1015447According to Mann-Whitney U analysis, there was a significant difference between the expression of DLL1 gene and the pathological grade Relationship between DLL1 expression and tumor characteristics in patients with OV According to Mann-Whitney U analysis, there was a significant difference between the expression of DLL1 gene and the pathological grade Whole transcriptome analysis of the clinical samples by RNA-sequencing identified a series of differential genes associated with OV (Fig. 2F). KEGG enrichment analysis showed that these genes were mainly enriched in the classical oncogenic pathways such as Notch signaling pathway, MAPK signaling pathway, and Wnt signaling pathway (Fig. 2G). Notably, as a core ligand of Notch pathway, DLL1 may affect the malignant phenotype of tumor cells by regulating downstream effector molecules of Notch pathway. The above results systematically revealed the aberrant expression characteristics of DLL1 in OV and its significant correlation with clinical prognosis, and also elucidated its potential mechanism of action through multi-omics analysis, which provided an important theoretical basis for the subsequent functional studies and the development of targeted therapeutic strategies. Result 2: DLL1 regulates OV malignant biological phenotypes To investigate the function of DLL1 in OV cells, we knocked down the expression of DLL1 in SK-OV-3 and A2780 cells lines using shRNA technology. Western blot analysis showed that the DLL1 protein expression level was significantly reduced in the shDLL1-transfected cells Figure S1C), suggesting the successful construction of the DLL1-knockdown cell lines. To assess the role of DLL1 in OV cell proliferation and clone formation, we performed functional analyses of SK-OV-3 and A2780 cell lines by CFDA-SE staining and clone formation assay as well as crystal violet staining. In the proliferative function analysis, CFDA-SE staining showed that DLL1 knockdown significantly inhibited the proliferative activity of OV cells. Compared with the sh-NC control group, the fluorescence intensity of the shDLL1 group showed a significant attenuation in both SK-OV-3 and A2780 cells (Fig. 3A), suggesting that down-regulation of DLL1 expression could block cell cycle progression. This effect was further corroborated in clone formation assays: quantitative analysis showed that shDLL1 reduced the number of clone formation by 103 (p < 0.001) and 99 (p < 0.001) in SK-OV-3 and A2780 cells, respectively (Fig. 3B). Crystal violet staining also showed a similar trend, with the knockdown group forming significantly fewer visible colonies than the control group (Figure S1D). Fig. 3DLL1 knockdown suppresses OV cell malignancy by inhibiting proliferation, clonogenicity, migration, and EMT while inducing cell death. A. Knockdown (shDLL1-1 or shDLL1-2) or no knockdown of DLL1 (sh-NC) in SK-OV-3 and A2780 cells, and cell activity was detected using CFDA-SE fluorescent staining; B. Colony formation assay, and count the number of clone formation in the experiment; C. Western blot detection and count the protein expression levels of E-cadherin, N-cadherin and Vimentin, and GAPDH was used as an internal reference protein; D. Transwell assay to count the cell migration fold change; E. Annexin V/PI double staining to detect the cells and count the cell death level DLL1 knockdown suppresses OV cell malignancy by inhibiting proliferation, clonogenicity, migration, and EMT while inducing cell death. A. Knockdown (shDLL1-1 or shDLL1-2) or no knockdown of DLL1 (sh-NC) in SK-OV-3 and A2780 cells, and cell activity was detected using CFDA-SE fluorescent staining; B. Colony formation assay, and count the number of clone formation in the experiment; C. Western blot detection and count the protein expression levels of E-cadherin, N-cadherin and Vimentin, and GAPDH was used as an internal reference protein; D. Transwell assay to count the cell migration fold change; E. Annexin V/PI double staining to detect the cells and count the cell death level To further explore the molecular mechanisms by which DLL1 regulates the EMT process, we examined the protein expression levels of epithelial-mesenchymal phenotypic transition markers. Western blot assay showed that DLL1 knockdown significantly altered the expression profiles of EMT-related proteins. In SK-OV-3 cells, shDLL1-1 treatment resulted in a 21% (p < 0.01) decrease in the epithelial marker E-cadherin, while the mesenchymal markers N-cadherin and cytoskeletal protein Vimentin were reduced by 64% (p < 0.001) and 53% (p < 0.001), respectively. A similar pattern of protein expression was observed in the A2780 cell line similar altered protein expression patterns were also observed (Fig. 3C), suggesting that DLL1 may promote tumor progression by affecting the EMT-related mesenchymal phenotype. The results of Transwell migration assay showed that shDLL1-1 treatment significantly impaired the invasive ability of OV cells. Quantitative statistics revealed that the migration fold change of SK-OV-3 cells in the shDLL1 group was reduced to 0.43 (p < 0.001), and that of A2780 cells was reduced to 0.25 (p < 0.001) (Fig. 3D), suggesting that DLL1 has an important role in maintaining the metastatic potential of tumor cells. Notably, cell death assays revealed that shDLL1-1 treatment induced programmed cell death. Annexin V/PI double staining analysis showed that the cell death rates of SK-OV-3 and A2780 cells in the knockdown group were increased to 13% (p < 0.001) and 14% (p < 0.001), respectively (Fig. 3E), suggesting that DLL1 may, through an anti-apoptotic mechanism promoting tumor cell survival. Result 3: DLL1 mediates ferroptosis resistance mechanisms through the Nrf2/GPX4 axis CCK-8 analysis revealed that DLL1 knockdown significantly reduced SK-OV-3 cell viability, while ferroptosis inhibitor Fer-1 and necroptosis inhibitor Nec-1 partially rescued this phenotype. In contrast, the caspase inhibitor z-VAD-FMK exhibited minimal protective effects (Fig. 4A). These data indicate that loss of DLL1 predominantly triggers ferroptosis-associated cell death. By Pearson correlation analysis, we examined the potential association between DLL1 expression levels and ferroptosis (Fig. 4B). The results showed a significant negative correlation between DLL1 and ferroptosis indicators (r=-0.10, p = 0.04). This statistical significance suggests that high expression of DLL1 may be accompanied by a reduced frequency of ferroptosis. Notably, despite the low absolute value of the correlation coefficient (|r|=0.10), its significance suggests that DLL1 may be involved in the regulation of ferroptosis through specific molecular mechanisms. Fig. 4DLL1 suppresses ferroptosis in OV via Notch-Nrf2/GPX4 axis to attenuate lipid peroxidation and promote tumor survival. A. Cell viability analysis of SK-OV-3 cells with DLL1 knockdown (shDLL1) treated with different ferroptosis inhibitor Fer-1, necroptosis inhibitor Nec-1 and caspase inhibitor z-VAD-FMK. B. Pearson correlation analysis of DLL1 expression levels with ferroptosis, with statistical significance value (p = 0.04) and a negative correlation coefficient (fPererson = -0.10); C. SK-OV-3 and A2780 cells knocked down (shDLL1) or not (shCtrl), or DLL1 knocked down with DLL1 recombinant protein (sh+DLL1) was supplemented at the same time, and the levels of ferroptosis-associated proteins GPX4 and Nrf2 were detected by western blot, along with the levels of DLL1 protein and the levels of GAPDH as an internal reference; D. Western blot detection of ferroptosis-related markers xCT and ACSL4 in SK-OV-3 and A2780 cells. E. ChIP-qPCR confirming direct transcriptional regulation of Nrf2 by Notch signaling. F. Dual-luciferase reporter assay validating that Notch activation enhances Nrf2 promoter activity. G. The activity of different subgroups of cells was detected using CFDA-SE fluorescent staining, and the relative proportion of CFDA-SE-positive cells was counted; H. C11-BODIPY 581/591 staining was used to detect the lipid peroxidation status of OV cells, the green fluorescent signal reflected the level of lipid peroxidation (the degree of oxidative stress), and the red fluorescent signal labeled the cells that did not undergo lipid peroxidation, and the level of Lipid-ROS production was counted DLL1 suppresses ferroptosis in OV via Notch-Nrf2/GPX4 axis to attenuate lipid peroxidation and promote tumor survival. A. Cell viability analysis of SK-OV-3 cells with DLL1 knockdown (shDLL1) treated with different ferroptosis inhibitor Fer-1, necroptosis inhibitor Nec-1 and caspase inhibitor z-VAD-FMK. B. Pearson correlation analysis of DLL1 expression levels with ferroptosis, with statistical significance value (p = 0.04) and a negative correlation coefficient (fPererson = -0.10); C. SK-OV-3 and A2780 cells knocked down (shDLL1) or not (shCtrl), or DLL1 knocked down with DLL1 recombinant protein (sh+DLL1) was supplemented at the same time, and the levels of ferroptosis-associated proteins GPX4 and Nrf2 were detected by western blot, along with the levels of DLL1 protein and the levels of GAPDH as an internal reference; D. Western blot detection of ferroptosis-related markers xCT and ACSL4 in SK-OV-3 and A2780 cells. E. ChIP-qPCR confirming direct transcriptional regulation of Nrf2 by Notch signaling. F. Dual-luciferase reporter assay validating that Notch activation enhances Nrf2 promoter activity. G. The activity of different subgroups of cells was detected using CFDA-SE fluorescent staining, and the relative proportion of CFDA-SE-positive cells was counted; H. C11-BODIPY 581/591 staining was used to detect the lipid peroxidation status of OV cells, the green fluorescent signal reflected the level of lipid peroxidation (the degree of oxidative stress), and the red fluorescent signal labeled the cells that did not undergo lipid peroxidation, and the level of Lipid-ROS production was counted To test this hypothesis, we designed a functional reply experimental system. western blot analysis showed that DLL1 knockdown significantly reduced the expression levels of Nrf2, a key protein in ferroptosis defense, and its downstream effector molecule GPX4. In SK-OV-3 cells, GPX4 and Nrf2 proteins decreased to 50% (p < 0.001) and 29% (p < 0.01) of the control, respectively, whereas supplementation with exogenous DLL1 restored them to 77% and 71% of the baseline levels (Fig. 4C). DLL1 knockdown significantly decreased xCT protein levels while increasing ACSL4 expression—two key molecular events that collectively amplify lipid peroxidation susceptibility (Fig. 4D). These findings further support the conclusion that DLL1 maintains ferroptosis resistance by preserving cystine import and suppressing lipid-PUFA activation. To clarify the upstream regulatory mechanism, ChIP-qPCR analysis revealed that Notch binds directly to the Nrf2 promoter region, confirming that Nrf2 is a transcriptional target of Notch signaling (Fig. 4E). A dual-luciferase reporter assay further validated that Notch activation robustly increases Nrf2 promoter activity (Fig. 4F). These data mechanistically link DLL1 to Nrf2 transcription via the canonical Notch signaling cascade. Further studies revealed that the anti-ferroptosis function of DLL1 significantly affected the proliferative homeostasis of tumor cells. CFDA-SE staining demonstrated that DLL1 knockdown resulted in a reduction of proliferation rate to 47% (p < 0.01) and 33% (p < 0.001) in SK-OV-3 and A2780 cells, respectively, which was significantly restored by exogenous DLL1 expression to 71% (p < 0.01) and 69% (p < 0.001) (Fig. 4G). This dose-dependent reversal effect suggests that DLL1 may indirectly inhibit the ferroptosis process by maintaining proliferative activity. To directly verify the regulatory effect of DLL1 on ferroptosis, we examined the dynamic changes of Lipid ROS, a key indicator of lipid peroxidation. The analysis showed that the percentage of Lipid ROS production increased dramatically in OV cells after knockdown of DLL1, and this change was reversed after concomitant supplementation of exogenous DLL1 (Fig. 4H): the percentage of Lipid ROS production decreased by 88% in SK-OV-3 cells (p < 0.001); and decreased by 56% in A2780 cells (p < 0.01). This result was highly consistent with the trend of Nrf2/GPX4 protein changes, revealing that DLL1 may inhibit the lipid peroxidation burst by up-regulating the antioxidant defense system. Result 4: DLL1 affects OV progression by regulating ferroptosis and mitochondrial homeostasis To define how DLL1 modulates ferroptosis in OV cells, we quantified core biochemical features of ferroptotic damage. DLL1 knockdown markedly increased malondialdehyde (MDA) accumulation (3.4-fold of control) and reduced intracellular glutathione (GSH) by ~ 87%, while elevated ferrous iron (Fe²⁺) content by 2.2-fold, indicating enhanced lipid peroxidation, antioxidant depletion, and iron overload (Fig. 5A). Reintroduction of DLL1 partially restored MDA, GSH, and Fe²⁺ levels, and similar biochemical patterns were observed in A2780 cells. To verify that these changes reflected bona fide ferroptosis, rescue experiments with the ferroptosis inhibitor Fer-1 were performed. Co-treatment with Fer-1 restored xCT, GPX4, and Nrf2 expression and suppressed ACSL4 accumulation in DLL1-deficient cells (Figure S2A), confirming that DLL1 loss primarily drives ferroptotic injury. Correspondingly, Fer-1 normalized MDA, GSH, and Fe²⁺ levels in both SK-OV-3 and A2780 cells (Figure S2B-C). We next assessed lipid peroxidation directly using C11-BODIPY 581/591 staining. DLL1 knockdown caused strong oxidation of the lipid probe, whereas Fer-1 markedly decreased lipid-ROS accumulation (Figure S2D). Fig. 5DLL1 maintains mitochondrial homeostasis and redox balance by attenuating oxidative stress and ferroptotic damage in OV cells. A. Knockdown (shDLL1) or no knockdown of DLL1 (shCtrl), or knockdown of DLL1 supplemented with exogenous DLL1 recombinant protein (sh+ DLL1) in SK-OV-3 or A2780 cells was performed to detect the MDA level, the GSH level, and the ferrous ion level; B. The mitochondria in different grouped cells were stained using Mitotracker-red; C. Mitortracker-green to observe mitochondrial localization and morphology, and to count the proportion of intact mitochondria (Inter) and fragmented mitochondria (Frag) DLL1 maintains mitochondrial homeostasis and redox balance by attenuating oxidative stress and ferroptotic damage in OV cells. A. Knockdown (shDLL1) or no knockdown of DLL1 (shCtrl), or knockdown of DLL1 supplemented with exogenous DLL1 recombinant protein (sh+ DLL1) in SK-OV-3 or A2780 cells was performed to detect the MDA level, the GSH level, and the ferrous ion level; B. The mitochondria in different grouped cells were stained using Mitotracker-red; C. Mitortracker-green to observe mitochondrial localization and morphology, and to count the proportion of intact mitochondria (Inter) and fragmented mitochondria (Frag) MitoTracker-Red staining demonstrated a ~ 61% reduction in mitochondrial activity in DLL1-knockdown SK-OV-3 cells (p < 0.001), which was restored to 77% of control levels following exogenous DLL1 supplementation (Fig. 5B). Notably, Fer-1 treatment also partially restored mitochondrial activity (Figure S2E), indicating that ferroptosis is a major contributor to the mitochondrial defects caused by DLL1 loss. Furthermore, mitochondrial morphology analysis using MitoTracker-Green revealed that DLL1 depletion resulted in pronounced mitochondrial fragmentation, blurred cristae, and loss of structural continuity in both OV cell lines (Fig. 5C). Quantification showed a marked shift from intact/interconnected mitochondria to fragmented forms. Both exogenous DLL1 restoration and Fer-1 treatment significantly increased the proportion of intact mitochondria while reducing fragmented structures (Figure S2F), demonstrating that DLL1 maintains mitochondrial structural integrity and protects against ferroptosis-associated mitochondrial injury. Taken together, these findings demonstrate that DLL1 suppresses ferroptosis by maintaining redox balance, limiting Fe²⁺-driven oxidative reactions, preserving mitochondrial activity, and preventing mitochondrial fragmentation. The ability of Fer-1 to reverse DLL1-knockdown phenotypes further confirms that ferroptosis is a key mechanism underlying DLL1-mediated regulation of OV cell fate. Result 5: DLL1 mediates cisplatin resistance in OV by modulating ferroptosis resistance To investigate the role of DLL1 in OV chemotherapy sensitivity, we examined the effect of DLL1 knockdown on cisplatin sensitivity in SK-OV-3 and A2780 cell lines. Dose-response curves were plotted by CCK-8 assay and showed that DLL1 knockdown significantly enhanced the sensitivity of both cell lines to cisplatin (Fig. 6A). In SK-OV-3 cells, the IC50 value of the DLL1 knockdown group (sh-DLL1-1) decreased from 2.0 µM to 1.5 µM compared with the control group (sh-NC) (p < 0.05). Similarly, the IC50 value of A2780 cells decreased from 2.5 µM to 0.9 µM (p < 0.05). This dose-dependent decrease in cell viability suggests that DLL1 knockdown can significantly increase the sensitivity of OV cells to cisplatin. Notably, under the treatment of high concentrations of cisplatin (5–10 µM), the cell viability of the DLL1 knockdown group decreased more significantly than that of the control group, suggesting that DLL1 may affect chemotherapy sensitivity through ferroptosis resistance mechanism. Changes in the viability of OV cells under cisplatin treatment were assessed by CFDA-SE staining (Fig. 6B). The results showed that cisplatin treatment (10 µM) significantly decreased cell viability (relative to CFDA-SE fluorescence intensity) in SK-OV-3 and A2780 cells. shDLL1 co-treatment with cisplatin further exacerbated the inhibition of cell viability by cisplatin, suggesting that DLL1 may play a role in OV cell resistance by affecting the cellular response to cisplatin. To investigate the mechanism of DLL1 in cisplatin resistance, we examined the effect of DLL1 knockdown on ferroptosis in SK-OV-3 cells under cisplatin (10 µM) treatment. The results showed that DLL1 knockdown significantly enhanced the effect of cisplatin-induced ferroptosis compared with the control (Fig. 6D). This was manifested by elevated levels of cellular lipid peroxidation and more significant GSH depletion. More notably, Fe2+ concentration showed a burst of increase in this model, and this iron overload phenomenon may lead to cascading amplification of toxic ROS through the Fenton reaction, thus forming a positive feedback loop to accelerate ferroptosis (Fig. 6E). Observation of mitochondrial morphology in OV cells by Mitotracker-green staining revealed (Fig. 6F) that cisplatin treatment resulted in the loss of mitochondrial cristae structure, a significant increase in the proportion of fragmented mitochondria (Frag), and a significant decrease in mitochondrial integrity (Inter) in OV cells. The mitochondrial damage was further exacerbated by concurrent knockdown of DLL1. Fig. 6DLL1 promotes cisplatin resistance via suppressing ferroptosis-driven lipid peroxidation and mitochondrial dysfunction in OV. A. SK-OV-3 or A2780 cells were treated with gradient concentrations of cisplatin, and cell viability was detected by CCK-8 assay, the curves were plotted and the IC50 was calculated; B. SK-OV-3 or A2780 cells were treated with 10 µM cisplatin for 48 h, and DLL1 (shDLL1-1) was knocked down at the same time. The effect of cisplatin on cell viability was quantified by CFDA-SE staining, and the relative fluorescence intensity was detected and statistically analyzed; C. Lipid peroxidation status of different subgroups of OV cells was detected by C11-BODIPY 581/591 staining, and the level of lipid-ROS production was statistically analyzed; D-E. The levels of MDA, GSH, and ferrous ions were detected in SK-OV-3 or A2780 cells; F. Staining of mitochondria from different subgroups of cells using Mitotracker-green, observation of mitochondrial localization and morphology, and counting of the proportion of intact mitochondria (Inter) and fragmented mitochondria (Frag) DLL1 promotes cisplatin resistance via suppressing ferroptosis-driven lipid peroxidation and mitochondrial dysfunction in OV. A. SK-OV-3 or A2780 cells were treated with gradient concentrations of cisplatin, and cell viability was detected by CCK-8 assay, the curves were plotted and the IC50 was calculated; B. SK-OV-3 or A2780 cells were treated with 10 µM cisplatin for 48 h, and DLL1 (shDLL1-1) was knocked down at the same time. The effect of cisplatin on cell viability was quantified by CFDA-SE staining, and the relative fluorescence intensity was detected and statistically analyzed; C. Lipid peroxidation status of different subgroups of OV cells was detected by C11-BODIPY 581/591 staining, and the level of lipid-ROS production was statistically analyzed; D-E. The levels of MDA, GSH, and ferrous ions were detected in SK-OV-3 or A2780 cells; F. Staining of mitochondria from different subgroups of cells using Mitotracker-green, observation of mitochondrial localization and morphology, and counting of the proportion of intact mitochondria (Inter) and fragmented mitochondria (Frag) Result 6: DLL1 enhances cisplatin resistance and ferroptosis resistance in OV by modulating the Nrf2/GPX4 signaling axis To elucidate the mechanism of DLL1 in cisplatin resistance in OV, we first constructed cisplatin-resistant cell models SK-OV-3-R and A2780-R. Western blot analysis showed that the expression level of DLL1 was significantly upregulated in drug-resistant cells (Cis-R) compared with cisplatin-sensitive cells (Cis-S). This upregulation occurred accompanied by increased levels of key factors of ferroptosis resistance synchronized elevation of Nrf2, GPX4 and xCT, while the ferroptosis indicator ACSL4 was downregulated (Figure S3). This finding further demonstrated that DLL1 might be involved in the formation of cisplatin resistance by regulating the ferroptosis resistance pathway. To test the hypothesis that targeting DLL1 promotes ferroptosis in cisplatin-resistant OV cell coat, we knocked down DLL1 expression in drug-resistant cell lines. The results showed that DLL1 knockdown significantly reduced the protein levels of GPX4 and Nrf2 (Fig. 7A), suggesting that DLL1 may regulate ferroptosis sensitivity through the Nrf2/GPX4 signaling axis. Further functional experiments confirmed that DLL1 knockdown (sh-DLL1-1) significantly sensitized drug-resistant cells to cisplatin: CFDA-SE staining showed a decrease in cell proliferation (Fig. 7B); and lipid peroxidation level assay showed an increase in Lipid ROS by 52% and 55%, respectively (p < 0.05) (Fig. 7C-D). In the ferroptosis key index assay, DLL1 knockdown resulted in increased MDA levels and decreased GSH content, along with increased intracellular free Fe²+ concentration (Fig. 7E). These changes were highly consistent with typical ferroptosis features, suggesting that DLL1 knockdown may reverse the drug-resistant phenotype by disrupting redox homeostasis and iron metabolism balance. This conclusion was further supported by mitochondrial functional analyses. MitoTracker Red staining showed that DLL1 knockdown decreased mitochondrial viability (Fig. 7F). MitoTracker Green staining combined with morphological analyses revealed an increase in the proportion of fragmented mitochondria (Frag) and a decrease in the proportion of intact mitochondria (Inter) in the knockdown group (Fig. 7G). This significant alteration in mitochondrial structure and function is consistent with the characterization of the execution phase of ferroptosis. Fig. 7DLL1 sustains cisplatin resistance in OV by reinforcing Nrf2/GPX4 antioxidant signaling and preserving mitochondrial homeostasis. A. Construct OV drug-resistant cell lines SK-OV-3-R and A2780-R, and then further knockdown (sh-DLL1-1 or sh-DLL1-2) or not (sh-NC) DLL1, western blot to detect the protein expression levels of DLL1, GPX4, and Nrf2, and GAPDH was the internal reference protein; B. Use CFDA-SE fluorescence staining to detect different subgroups of cell activity and count the relative CFDA-SE-positive cell proportion; C-D. Lipid peroxidation status of OV cells was detected by C11-BODIPY 581/591 staining, and the level of Lipid-ROS production was counted; E. The levels of MDA, GSH, and ferrous ions were detected in different subgroups of drug-resistant OV cells; F. The expression levels of GAPDH and Nrf2 proteins in different subgroups of cells were detected by Mitortracker-red staining. Mitortracker-red to stain the mitochondria of different grouped cells and count the relative proportion of positive cells; G. Observation of mitochondrial localization and morphology using Mitotracker-green to count the proportion of intact mitochondria (Inter) and fragmented mitochondria (Frag) DLL1 sustains cisplatin resistance in OV by reinforcing Nrf2/GPX4 antioxidant signaling and preserving mitochondrial homeostasis. A. Construct OV drug-resistant cell lines SK-OV-3-R and A2780-R, and then further knockdown (sh-DLL1-1 or sh-DLL1-2) or not (sh-NC) DLL1, western blot to detect the protein expression levels of DLL1, GPX4, and Nrf2, and GAPDH was the internal reference protein; B. Use CFDA-SE fluorescence staining to detect different subgroups of cell activity and count the relative CFDA-SE-positive cell proportion; C-D. Lipid peroxidation status of OV cells was detected by C11-BODIPY 581/591 staining, and the level of Lipid-ROS production was counted; E. The levels of MDA, GSH, and ferrous ions were detected in different subgroups of drug-resistant OV cells; F. The expression levels of GAPDH and Nrf2 proteins in different subgroups of cells were detected by Mitortracker-red staining. Mitortracker-red to stain the mitochondria of different grouped cells and count the relative proportion of positive cells; G. Observation of mitochondrial localization and morphology using Mitotracker-green to count the proportion of intact mitochondria (Inter) and fragmented mitochondria (Frag) Result 7: Combined cisplatin and RSL3 treatment overcomes DLL1-mediated ferroptosis resistance through GPX4 inhibition To elucidate the functional mechanism of DLL1 in the regulation of ferroptosis in OV cells, we overexpressed DLL1 in SK-OV-3 and A2780 cells and examined its effect on GPX4, a key regulator of ferroptosis. Western blot analysis showed that either cisplatin or the GPX4 inhibitor RSL3 alone significantly reduced the expression level of GPX4, while the combined treatment resulted in a GPX4 expression to be further downregulated (Figure S4A). This finding suggests that DLL1 may affect cellular ferroptosis sensitivity by regulating GPX4 expression. In the cell viability assay, CFDA-SE staining showed that cisplatin or RSL3 treatment alone decreased the viability of cells overexpressing DLL1, whereas co-treatment resulted in a significant decrease in cell viability to below 52% (SK-OV-3: 51%; A2780: 49%) (Fig. 8A). This synergistic effect was further validated in the lipid peroxidation level assay: co-treatment increased Lipid ROS levels by 191% and 199%, respectively, compared to treatment alone (Fig. 8B). The ferroptosis key indicator assay showed that co-treatment significantly increased MDA levels and GSH depletion, along with an increase in intracellular free Fe²+ concentration (Figure S4B). Mitochondrial morphology analysis further confirmed that co-treatment resulted in a decrease in mitochondrial integrity (Inter) and an increase in fragmentation ratio (Frag) (Fig. 8C-F). Fig. 8Combination of cisplatin and ferroptosis inducer RSL3 overcomes DLL1-mediated resistance via synergistic activation of lipid peroxidation and mitochondrial dysfunction. A. SK-OV-3 and A780 cells were transfected with DLL1 overexpression plasmid (oe-DLL1) or empty vector (Vector), then treated with cisplatin (Cis) alone, RSL3 alone, or their combination. Cell viability was assessed by CFDA-SE staining, and relative fluorescence intensities were quantified; B. C11-BODIPY 581/591 staining was used to detect the lipid peroxidation status of OV cells, and the level of Lipid-ROS production was counted; C-F. Mitochondrial localization and morphology were observed using Mitotracker-green, and the ratio of intact mitochondria (Inter) to fragmented mitochondria (Frag) was counted; G. The lipid peroxidation status of the cisplatin-resistant cell lines (SK-OV-3-R or A2780-R) was detected by C11-BODIPY 581/591 staining, OV cells was treated with cisplatin treatment alone (Cis), or RSL3 alone, or simultaneous cisplatin and RSL3 co-treatment Combination of cisplatin and ferroptosis inducer RSL3 overcomes DLL1-mediated resistance via synergistic activation of lipid peroxidation and mitochondrial dysfunction. A. SK-OV-3 and A780 cells were transfected with DLL1 overexpression plasmid (oe-DLL1) or empty vector (Vector), then treated with cisplatin (Cis) alone, RSL3 alone, or their combination. Cell viability was assessed by CFDA-SE staining, and relative fluorescence intensities were quantified; B. C11-BODIPY 581/591 staining was used to detect the lipid peroxidation status of OV cells, and the level of Lipid-ROS production was counted; C-F. Mitochondrial localization and morphology were observed using Mitotracker-green, and the ratio of intact mitochondria (Inter) to fragmented mitochondria (Frag) was counted; G. The lipid peroxidation status of the cisplatin-resistant cell lines (SK-OV-3-R or A2780-R) was detected by C11-BODIPY 581/591 staining, OV cells was treated with cisplatin treatment alone (Cis), or RSL3 alone, or simultaneous cisplatin and RSL3 co-treatment To verify the generalizability of this phenomenon in drug-resistant cells, we performed parallel experiments in cisplatin-resistant cell models (SK-OV-3-R and A2780-R). The results showed that co-treatment increased Lipid ROS levels in resistant cells compared to treatment alone (Fig. 8G). Ferroptosis indicator assays showed significantly higher MDA levels, increased GSH depletion, and increased Fe²+ concentrations (Figure S4C). These results were highly consistent with cisplatin-sensitive cells, suggesting that the DLL1/GPX4 regulatory axis also plays an important role in drug-resistant cells. Result 8: DLL1 overexpression promotes tumor growth and therapy resistance, whereas combined cisplatin/RSL3 treatment overcomes DLL1-driven ferroptosis resistance in vivo To verify the functional role of DLL1 in OV in vivo, we successfully constructed DLL1 overexpressing SK-OV-3 and A2780 cell lines (Fig. 9A). Western blot assay showed that the level of DLL1 protein in the cells of the overexpressing group (DLL1) was significantly higher than that of the control group (Vector), confirming the success of the construction of cell models. In in vivo experiments, we inoculated DLL1 overexpressing SK-OV-3 cells subcutaneously in nude mice to construct a hormonal tumor model. The results showed that DLL1 overexpression significantly promoted tumor growth: 24 days after inoculation, the tumor volume of the DLL1 group increased by a factor of 1 compared with the control group (Fig. 9B). Tumor weight assay showed that the average tumor weight in the DLL1 group reached 1.44 g, which was significantly higher than that in the control group of 0.86 g (p < 0.01) (Fig. 9C-D). To assess the effect of DLL1 on treatment response, we treated with cisplatin monotherapy (DLL1 + Cis) and cisplatin combined with ferroptosis inducer RSL3 (DLL1 + Cis+RSL3), respectively. The results showed that single-agent treatment resulted in a reduction in tumor volume compared with the DLL1 group, whereas the combination treatment further inhibited tumor growth and resulted in a further reduction in tumor volume (Fig. 9B). Tumor weight assay showed that the average tumor weight decreased to 0.28 g in the combination therapy group, which was significantly lower than that of 0.86 g in the monotherapy group (p < 0.001) (Fig. 9D). Fig. 9DLL1-driven tumor growth and cisplatin resistance are reversed by co-targeting ferroptosis In Vivo, suppressing proliferation via Ki-67 downregulation. A. DLL1 was overexpressed in OV cell lines SK-OV-3 and A2780, and changes in DLL1 protein levels were detected by western blot; B. SK-OV-3 cells overexpressing DLL1 were inoculated subcutaneously in nude mice, and the control group was inoculated with wild-type SK-OV-3 cells, and for the overexpression group, cisplatin-treated treatment (Cis) or cisplatin combined with RSL3 treatment (Cis+ RSL3), and the subcutaneous tumor volume of mice was tracked for 24 days after loading; C-D. Mice were executed after 24 days of loading, and the tumors were collected and counted for tumor weight; E. H&E staining and Ki-67 staining were performed on the collected tumors, and the relative proportion of Ki-67-positive cells was counted DLL1-driven tumor growth and cisplatin resistance are reversed by co-targeting ferroptosis In Vivo, suppressing proliferation via Ki-67 downregulation. A. DLL1 was overexpressed in OV cell lines SK-OV-3 and A2780, and changes in DLL1 protein levels were detected by western blot; B. SK-OV-3 cells overexpressing DLL1 were inoculated subcutaneously in nude mice, and the control group was inoculated with wild-type SK-OV-3 cells, and for the overexpression group, cisplatin-treated treatment (Cis) or cisplatin combined with RSL3 treatment (Cis+ RSL3), and the subcutaneous tumor volume of mice was tracked for 24 days after loading; C-D. Mice were executed after 24 days of loading, and the tumors were collected and counted for tumor weight; E. H&E staining and Ki-67 staining were performed on the collected tumors, and the relative proportion of Ki-67-positive cells was counted Histopathological analysis further confirmed the above results.H&E staining showed that the tumor cells in the DLL1 group showed significant heterogeneity: the cells were loosely arranged, the nucleoplasmic ratio was enlarged, and the nucleoli were obvious (Fig. 9E). In the treatment groups, especially in the combination therapy group, the tumor cell heterogeneity was significantly improved, and the cell arrangement tended to be tight.Ki67 immunohistochemistry showed that the relative Ki67 positivity rate in the DLL1 group was significantly higher than that in the control group. Cisplatin monotherapy decreased the relative Ki-67 positivity rate, which was further reduced by combination therapy (Fig. 9E). These in vivo results confirm that DLL1 overexpression promotes OV growth by modulating ferroptosis resistance, and targeting DLL1 in combination with ferroptosis inducers significantly enhances the therapeutic effect of cisplatin. This finding provides an important preclinical rationale for the development of combination therapeutic regimens based on ferroptosis sensitization in OV. Discussion In this study, we systematically elucidated the pivotal role of the transmembrane ligand DLL1 in the malignant progression and cisplatin resistance of OV (OV) and revealed a novel molecular mechanism of its regulation of ferroptosis sensitivity through the Notch-Nrf2/GPX4 signaling axis. The main findings can be summarized in the following four aspects: First, DLL1 was significantly highly expressed in OV tissues and cell lines, and was significantly inversely associated with patient overall survival. Second, knockdown of DLL1 inhibited cancer cell proliferation, anti-cell death and metastasis. Meanwhile, this study revealed the potential mechanism of DLL1 and ferroptosis resistance. DLL1 activated the Nrf2-dependent antioxidant defense system, upregulated GPX4, inhibited lipid peroxidation, and was associated with the maintenance of mitochondrial structural integrity. Finally, this study found that targeting DLL1 partially reversed cisplatin resistance and significantly enhanced chemosensitivity in OV cells by inducing iron overload and lipid ROS burst. These results provide preclinical evidence supporting DLL1 as a potential therapeutic target and a candidate strategy for reversing chemotherapy resistance in OV. Temporal and spatial regulation properties of DLL1 and Notch pathway This study is the first to correlate the ligand function of DLL1 with the regulation of oxidative stress downstream of Notch. Existing studies have shown that Notch signaling plays a key role in tumor stemness maintenance and cisplatin resistance [27, 35, 36], but most focus on receptors (Notch1-4) or intracellular effector molecules. In this study, we found that DLL1 activates Notch via an autocrine/paracrine form that triggers Nrf2/GPX4 signaling, which expands the functional boundaries of Notch ligands. Notably, DLL family members (DLL1/4, Jagged1/2) may possess different downstream targeting preferences: while Jagged1 has been shown to regulate cell death through the PI3K-Akt pathway in OV [37], DLL1 mediates ferroptosis resistance through GPX4, and this sorting specificity provides a potential direction for targeted intervention. Cross-linking mechanisms of ferroptosis and chemotherapy resistance Recent studies have revealed that ferroptosis resistance is an important mechanism of chemotherapeutic resistance [38, 39], mainly mediated through antioxidant molecules such as GPX4 and xCT. In the present study, we found compensatory activation of the DLL1-Nrf2/GPX4 axis in cisplatin-resistant OV (Cis-R) to form a dynamic antioxidant barrier. The unique value of this study is to reveal the regulation of cellular iron homeostasis by Notch signaling: knockdown of DLL1 leads to a surge in intracellular Fe²+ concentration, which not only promotes lipid ROS accumulation through the Fenton reaction, but also potentially enhances iron storage through endocytosis regulation by membrane transferrin receptors such as TFR1. This finding extends the traditionally recognized role of Notch pathway as an oncogene to the regulation of redox homeostasis, providing a new dimension for understanding drug resistance mechanisms. Notably, in resistant cells with high DLL1 expression, DLL1 counteracts the toxic stress by maintaining GSH synthesis even when cisplatin inhibits glutathione reductase activity [40], which may partially explain why cisplatin monotherapy has limited efficacy in the clinic. It should be noted that the ferroptosis-promoting phenotypes observed after DLL1 knockdown and those observed in DLL1-overexpressing cells exposed to cisplatin plus RSL3 represent two mechanistically distinct experimental contexts rather than contradictory findings. DLL1 knockdown removes an endogenous Notch-Nrf2/GPX4-dependent antioxidant defense, directly sensitizing OV cells to lipid peroxidation, iron accumulation, and ferroptotic injury. Conversely, DLL1 overexpression creates a ferroptosis-resistant cellular state, though this protection partially relies on GPX4 activity. Thus, RSL3’s pharmacological inhibition of GPX4 can bypass DLL1-mediated protection, especially under cisplatin-induced oxidative stress. The enhanced ferroptotic response in DLL1-overexpressing cells treated with cisplatin plus RSL3 indicates that the DLL1-driven resistant phenotype is functionally vulnerable to downstream GPX4 blockade, rather than suggesting that DLL1 overexpression itself sensitizes cells to ferroptosis. This interpretation supports the therapeutic rationale that targeting ferroptosis-execution nodes can overcome DLL1-associated chemoresistance. The role of mitochondrial homeostasis in the regulation of ferroptosis The study revealed that DLL1 deficiency led to mitochondrial fragmentation and triggered ferroptosis in synergy with GPX4 downregulation by Mitotracker-green staining. While recent studies have focused on the role of lipid metabolizing enzymes such as ACSL4 and LOX in ferroptosis [41], the present study revealed that DLL1 enhances the survival of cancer cells by maintaining the integrity of mitochondria, suggesting that abnormalities in the structure and function may represent a critical step in the execution of ferroptosis. This regulation may be related to mitochondrial kinetic proteins; however, the molecules were not detected in this study, and the real-time regulation of mitochondrial membrane potential by DLL1 needs to be explored in the future in conjunction with ultra-high-resolution imaging. Clinical translational potential of DLL1-targeted therapy In our in vivo model, combining cisplatin with the ferroptosis inducer RSL3 resulted in a 65% shrinkage of DLL1 overexpressing tumors (p < 0.001), an effect that was significantly superior to conventional chemotherapeutic regimens (e.g., bevacizumab in combination with chemotherapy). These preclinical results suggest that targeting DLL1 may help overcome cisplatin resistance in experimental models. It is worth mentioning that monoclonal antibodies to DLL1 have shown safety in colon cancer models, and the therapeutic strategy in this study provides a theoretical reference for its combination. However, the possible normal tissue toxicity (e.g., liver injury) caused by excessive inhibition of ferroptosis needs to be guarded against, and targeted delivery systems responsive to the tumor microenvironment need to be developed in the future. Although this study systematically revealed the regulatory network of DLL1, the following limitations still exist: First, the clinical validation in this study was only based on a small sample cohort, which needs to be expanded to multicenter and ethnically diverse samples to verify the generalizability. Meanwhile, the mechanistic integrity of this study needs to be expanded. How Notch signaling specifically activates Nrf2 is still unclear, and the direct regulatory relationship needs to be verified by means of ChIP-seq and other means. Second, the therapeutic time window needs to be further explored to explore, and the impact of the timing of ferroptosis induction (e.g., sequential use prior to neoadjuvant chemotherapy) on the efficacy remains to be elucidated. Third, the in vivo experiments in this study have no independent control groups such as shDLL1 monotherapy or ferroptosis inducer RSL3 monotherapy were included. In addition, future studies should construct xenograft or orthotopic tumor models using cisplatin-resistant OV cells, such as SK-OV-3-R or A2780-R cells, to further validate whether targeting DLL1 can overcome ferroptosis resistance and cisplatin resistance in vivo. Future studies will include these additional controls and expand the animal sample size to provide more comprehensive evidence for clinical translation. In summary, this study expands the research paradigm of cisplatin resistance mechanisms in OV from “gene mutation-driven” to “redox homeostasis remodeling”, with three potential translational implications: First, the study suggests a revised mechanistic model of drug resistance, shifting from the traditional “transporter‑mediated drug efflux” paradigm to highlighting the DLL1-GPX4 axis-mediated antioxidant defense as a key contributor to cisplatin resistance. This perspective may help explain why olaparib remains effective in some platinum‑resistant patients [42] (potentially through synergistic ferroptosis induction). Second, the study proposes innovative combination therapy targets. Experimental validation shows that co-administration of DLL1 inhibitors with ferroptosis inducers can effectively reverse drug resistance. This “metabolic vulnerability targeting” strategy represents a potential next-generation therapeutic approach following immune checkpoint inhibitors. Third, the findings advance diagnostic biomarker development. DLL1 expression levels serve as a predictive biomarker for cisplatin efficacy, with dynamic monitoring potentially enabling real-time clinical regimen adjustments. Collectively, this work not only elucidates a novel molecular regulatory axis but also reveals the dynamic interplay among “spatial signaling, metabolic remodeling, and therapeutic resistance”, establishing a transformative theoretical framework and experimental basis for further preclinical research in precision OV therapy. Innovation This study identifies DLL1 as a clinically relevant driver of cisplatin resistance and reveals a previously unrecognized DLL1–Notch–Nrf2 transcriptional axis that coordinately regulates ferroptosis defense in OV. By integrating multi-omics profiling, clinical cohort validation, and mechanistic dissection, we demonstrate that Notch directly governs Nrf2-dependent GPX4 activation, establishing a new regulatory layer that links developmental signaling to ferroptosis resistance. The work further uncovers ferroptosis as the dominant death pathway underlying cisplatin tolerance through multi-parametric rescue and ultrastructural validation. These findings not only expand the mechanistic landscape of ferroptosis regulation but also highlight DLL1 as a promising biomarker and therapeutic target for improving chemoresistance in OV. Materials and methods Electronic laboratory notebook was not used Data resources and RNA-sequencing In this study, we integrated multi-omics data to systematically analyze the expression characteristics and functional mechanisms of DLL1 in OV. Gene set source and preprocessing: OV transcriptome datasets GSE6008 (containing normal ovarian tissue and primary OV samples), GSE148003 (cisplatin-resistant and sensitive samples), and GSE201600 (paired samples before and after neoadjuvant chemotherapy) were downloaded from the GEO database, and the raw data were preprocessed by R language (v4.2.1), including background correction (RMA algorithm), normalization (quantile normalization) and batch effect correction (ComBat). Differential expression analysis was performed using the limma package (v3.54.0) to compare the differences in DLL1 expression in tumor/normal tissues, drug-resistant/sensitive groups, and before and after treatment, with the thresholds set at |log2FC|>1 and FDR < 0.05. Prognostic analyses were performed by dividing the patients into high-/low-expression groups based on the median DLL1 expression, and the Kaplan-Meier method was used to Survival curves were plotted, and log-rank test was used to assess the difference in overall survival (OS) between groups (R package survival, v3.4-0). Ferroptosis correlation analysis: the core set of ferroptosis genes was extracted from the FerrDb database (v2.0), and the association between DLL1 expression and ferroptosis scores (based on gene set enrichment scores) was calculated by the Pearson’s correlation coefficient (R languagepsych package, v2.2.9). RNA-seq experiments and bioinformatics analysis: 3 cases of OV and paired paracarcinoma tissues (pathologically confirmed) were collected, total RNA was extracted by TRIzol method, and strand-specific libraries were constructed (Illumina NovaSeq 6000, PE150) after passing the quality control (RIN > 8). Raw sequencing data were compared to the human reference genome GRCh38 using HISAT2 (v2.2.1) and gene counts were performed by featureCounts (v2.0.3) after FastQC quality control. Differential gene screening was performed using DESeq2 (v1.38.0) with the threshold set at |log2FC|>1 and padj < 0.01. Functional enrichment analysis: KEGG pathway and GO (biological process) enrichment analysis (FDR < 0.05) was performed on differential genes by clusterProfiler (v4.6.0). Construction of cell lines and cisplatin-resistant cell lines The OV cell lines used in this study included IOSE80, HEY, A2780 and SK-OV-3, which were purchased from ATCC. IOSE80 is a normal ovarian epithelial cell line, and HEY, A2780 and SK-OV-3 are commonly used OV cell lines. Cisplatin-resistant cell lines A2780-R and SK-OV-3-R were constructed by long-term culture in medium containing increasing concentrations of cisplatin: First, A2780 or SK-OV-3 cells were seeded in culture flasks containing standard medium, and after the cells grew to 70%-80% confluence, a low concentration of cisplatin (0.1 µM) was added and cultured for 3–5 days, and after the cells recovered, a higher concentration of cisplatin (0.5 µM) was added. The cisplatin-resistant cell lines were established by increasing the cisplatin concentration several times until the cells survived and were able to grow continuously under the culture conditions of 10 µM cisplatin. The cisplatin resistance of the resistant cell lines was verified by the cell growth inhibition rate detected by CCK-8 kit. Clinical samples In this study, tumor samples (T) and corresponding paracancerous samples (N) were collected from three OV patients. All samples were surgically resected at Obstetrics & Gynecology Hospital of Fudan University, and the process of collecting samples was approved by the ethical committee (Approval number: kyy2026-036), which complied with the ethical requirements. None of the patients received any treatment (including chemotherapy or radiotherapy) before sample collection. Clinical information of all patients, such as age, stage, grading, and treatment history, was collected and documented in detail through medical record review. Tumor samples and paracancerous tissues were selected based on the criteria that the tumor tissue was OV tissue and the paracancerous tissue was normal ovarian tissue at least 2 cm from the tumor. Western blot To detect the levels of GAPDH, DLL1, E-cadherin, N-cadherin, Vimentin, GPX4, and Nrf2 proteins in the samples, we analyzed them using western blot. We used western blot analysis. Protein samples were extracted repeatedly by freeze-thaw cycles in RIPA lysis buffer. We used a Pierce BCA protein quantification kit (A65453, Thermo Scientific) to determine protein concentration. Afterwards, 10–15 µg of protein was loaded onto an 11% SDS-PAGE gel for electrophoresis. Proteins were transferred to nitrocellulose membranes using the semi-dry transfer method. The membranes were blocked with 5% milk solution and subsequently incubated with specific primary antibodies: GAPDH (60004-1-Ig, Proteintech), DLL1 (28544-1-AP, Proteintech), E-cadherin (20874-1-AP, Proteintech), N-cadherin (22018-1-AP, Proteintech), Vimentin (60330-1-Ig, Proteintech), GPX4 (67763-1-Ig, Proteintech), Nrf2 (16396-1-AP, Proteintech), xCT(26864-1-Ig, Proteintech), ACSL4 (22401-1-Ig, Proteintech). The membranes were washed five times and then incubated for 1 h at room temperature with the corresponding secondary antibodies: rabbit-origin secondary antibody (SA00001-2, Proteintech), mouse-origin secondary antibody (SA00001-1, Proteintech). Protein bands were visualized by ECL solution (32106, Thermo Scientific) and detected using an automated chemiluminescence imaging system; ImageJ software was used to analyze and quantify band intensity. Immunohistochemistry staining Collected clinical tumor specimens and mouse tumor tissues were processed for standard paraffin embedding. After paraffin embedding, tissue samples were sectioned at 4 μm thickness and mounted on slides, which underwent routine deparaffination and rehydration steps. For clinical tumor samples, IHC staining was performed using anti-DLL1 antibody (28544-1-AP, Proteintech): sections were first treated with antigen repair to release cross-links and increase the binding capacity of the antibody. The sections were then treated with 3% hydrogen peroxide solution to remove endogenous peroxidase activity. Anti-DLL1 antibody was then added for overnight incubation at 4 °C. The next day, the sections were incubated with secondary antibody (SA00001-2, Proteintech) and stained with DAB color development system, and finally the sections were restained with hematoxylin (H9627-25G, Sigma-Aldrich). All IHC sections were observed under a microscope and photographed for documentation. For tumor tissues from mice after tumor loading, immunohistochemical staining with Ki67 antibody (27309-1-AP, Proteintech) was performed to detect the proliferative activity of tumor cells, and the relative percentage of Ki67-positive cells was counted. CFDA-SE staining CFDA-SE staining was used to detect the proliferation of OV cells. First, OV cells were inoculated in 6-well plates, and after the cells grew to 70%-80% confluence, the cells were washed twice with serum-free medium. Subsequently, CFDA-SE solution (5 µM) (HY-D0938, MCE) was added to the cells and incubated for 30 min at room temperature. At the end of the incubation, the cells were washed twice with serum-free medium to remove unbound CFDA-SE. The fluorescence signal of CFDA-SE staining was detected and counted using fluorescence microscopy. Clone formation experiments and crystal violet staining Clone formation assays were used to assess the long-term proliferation and viability of OV cells. First, OV cells were inoculated in 6-well plates and cultured for 14 days, during which the medium was changed every 2–3 days. When the clones grew to a visible size, the cells were gently washed using PBS. Subsequently, the cells were stained using crystal violet solution (0.5%) (HY-B0324A, MCE) for 10 min and washed with water after staining. Clones were observed and counted using a microscope; the number of clones and clone size reflected the proliferation and viability of the cells. A higher number of clones indicated that the cells had a strong proliferative potential. Transwell experiment First, cells were seeded in medium containing 10% FBS (A5256701, Thermo Scientific) in the upper chamber of the Transwell and incubated for 24 h to allow cell attachment. Medium containing 20% FBS was added to the lower chamber as a chemoattractant. After 24 h, cells in the upper and lower chambers were washed with PBS. Unmigrated cells in the upper chamber were gently wiped away using a cotton swab, and then cells in the lower chamber were fixed with methanol and stained with crystal violet for 10 min. The number of migrated cells in the lower chamber was observed and counted with a microscope, and the migration fold change was counted. Annexin V/PI double staining Annexin V/PI double staining was used to detect cell death in OV cells. Cells were treated with Annexin V-FITC staining solution and PI solution according to the instructions of Annexin V/PI detection kit (HY-K1073, MCE). After staining, the apoptotic status of the cells was analyzed using flow cytometry. C11-BODIPY 581/591 staining method The C11-BODIPY 581/591 staining (95978, Cell Signalling) method was employed to assess lipid peroxidation status in OV cells, where the green fluorescent signal corresponded to lipid peroxidation levels (serving as an indicator of oxidative stress), while the red fluorescent signal identified non-peroxided cells, allowing quantification of Lipid-ROS production. Measurement of MDA, GSH and ferrous ion levels Standardized experiments were performed according to the manufacturer’s instructions. MDA and levels were measured using the Lipid Peroxidation MDA Assay Kit (S0131M, Beyotime). GSH levels in sample cells were measured using the GSH Assay Kit (S0053, Beyotime). Ferrous ion levels were measured by Iron Assay Kit (MAK025, Sigma-Aldrich). Mitochondrial staining Cellular mitochondria were stained using Mitotracker-red kit (9082, Cell Signalling) or Mitotracker-deep red kit (8878, Cell Signalling) depending on the manufacturer’s protocol look. For mitochondria after Mitochondira-red staining, the relative percentage of positivity was counted. For mitochondria after Mitochondira-green staining, the proportion of intact mitochondria (Inter) and fragmented mitochondria (Frag) was counted. CCK-8 experiment OV cells are inoculated into 96-well plates and treated with a gradient concentration of cisplatin (0–10 µM). 10 µL of CCK-8 reagent (96992, Sigma-Aldrich) was added to each well and incubated for 1 h at 37 °C and 5% CO2. Absorbance was measured at 450 nm using a microplate reader. Construction of a nude mouse tumor model The animals used in this study were 6-week-old female BALB/c nude mice purchased from National Seed Center for Rodent Laboratory Animals. All mice were acclimatized for one week prior to the experiments and maintained under specific environmental conditions during the experiments, with a temperature of 22 ± 2 °C, relative humidity of 50%-60%, and a 12-hour light-dark cycle. Tumor cells were inoculated with 1 × 106 SK-OV-3 cell suspension and injected into the right axillary region of mice by subcutaneous injection. Each mouse was injected with approximately 180 µL of cell suspension. After tumor formation, the diameters of the tumors were measured periodically using calipers to calculate the tumor volume to assess tumor growth. Tumor volume measurements were performed every 3 days for 24 days. At the end of the experiment, mice were euthanized, after which the tumor tissues were collected, photographs were taken to record the size of the tumors, and the weight of the tumors was measured with a precision balance. All experimental procedures were performed in strict compliance with the ethical requirements for animal experiments. H&E staining Collected mouse tumor tissues were processed for standard paraffin embedding and sectioned for standard H&E staining. The sections were first stained with hematoxylin (H9627-25G, Sigma-Aldrich) for 5 min, followed by staining with acidic alcohol, and finally stained with eosin (HT110116, Sigma-Aldrich) for 2–3 min. The stained sections were washed under running water, dehydrated, visualized and sealed. Sections were observed microscopically and morphological changes in the tumor tissue were recorded at different magnifications, including the size and shape of tumor cells, nucleoplasmic ratio and any pathological changes. ChIP-qPCR assay ChIP was performed to assess Notch1 binding to the Nrf2 promoter. Cells were cross-linked, sonicated (200–500 bp fragments), and immunoprecipitated with anti-Notch1 (ab52627. Abcam) or IgG control. Purified DNA was analyzed by qPCR. Enrichment was normalized to input DNA. Dual-luciferase reporter assay The Nrf2 promoter was cloned into pGL3-basic. Cells were co-transfected with pGL3-Nrf2 and pRL-TK using Lipofectamine 3000. Luciferase activity was measured using Promega Dual-Luciferase System and normalized as Firefly/Renilla. Data processing and analysis Data are expressed as mean ± standard deviation (SD). Statistical significance was calculated using GraphPad Prism (v10.1.2). Comparisons of three or more means were analyzed using one-way ANOVA with two-way comparisons by Tukey’s method, whereas two means were compared using the unpaired Student t-test; *p < 0.05, **p < 0.01 and ***p < 0.001 was considered statistically significant. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2