KELLY and SK-N-BE (2) (MYCN amplified), SK-N-AS and SK-N-FI (MYCN non-amplified) NB cell lines were employed in this study. All cell lines were cultured in complete media, consisting of RPMI 1640 (Gibco) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin, at 37°C and 5% CO₂. The natural killer cell line NK-92, generously provided by Prof. Kerry Campbell, was cultured in alpha-MEM containing ribo- and deoxyribonucleosides, supplemented with 12.5% heat inactivated (HI) horse serum (Sigma-Aldrich), 12.5% HI fetal calf serum (FCS), 2.2 g/L sodium bicarbonate, 0.2 mM inositol (Sigma-Aldrich), 2 mM L-glutamine, 0.1 mM 2-mercaptoethanol (Life Technologies), 0.02 mM folic acid (Sigma-Aldrich), and 200 U/mL recombinant IL-2 (Chiron). All cell lines were cultured under standard conditions and routinely tested for the presence of mycoplasma, whereas NB cell line identities were verified by SNP microarray analysis (33).

KELLY, SK-N-AS, SK-N-BE (2), and SK-N-FI NB cell lines were seeded in 96-well plates at a density of 8000 cells/well. The next day, cells were treated with different concentrations of the System Xc inhibitor Erastin (Selleckchem) (range 0.05-10.0 µM). For combination experiments, cells were treated with either Erastin (Selleckchem), the ferroptosis inhibitor Liproxstatin-1 (Selleckchem), or a combination of both. Cellular viability was assessed at 72 h with resazurin assay, with cells incubated with 44 µM resazurin (Sigma Aldrich) for 2-3 h, followed by fluorescence (λex/em=560/590nm) quantification in a TECAN microplate reader.

Cells (1.0 × 10⁵) were seeded in 12-well plates in triplicate and treated with either 1 or 5 µM Erastin with DMSO as a control for 14 days. Plates were washed with PBS and air-dried, then fixed with methanol for 20 minutes. Cells were stained with 0.2% crystal violet in 20% ethanol for 20 minutes, followed by three brief rinses in water. Plates were then scanned using Toshiba Studio 2505AC.

Proteins from respective cell lines and control cells were extracted using cell lysis buffer (Cell Signaling) supplemented with proteinase and phosphatase inhibitors. Protein amount was quantified using the Pierce BCA Protein assay kit (Thermo Fisher Scientific). The exact amount of total protein sample (20 µg) was separated from each sample using SDS-PAGE and transferred onto a 0.2-µm PVDF membrane. The blots were probed overnight at 4°C with the following antibodies: ULBP1 (#703182, 1:500) from Thermo Fisher, GAPDH (#2118, 1:10,000), and ATF4 (#11815, 1:1,000), which were purchased from Cell Signaling. Horseradish peroxidase–conjugated secondary antibodies, goat anti-mouse immunoglobulin G (IgG) (#sc-2748, 1:5000), and mouse anti-rabbit IgG (#sc-2357, 1:5000) were purchased from Santa Cruz. Protein levels were quantified using ImageJ software.

ATF4 was silenced in NB cells (KELLY and SK-N-AS) using Silencer Select siRNAs (Invitrogen) according to the manufacturer’s instructions. Cells were seeded in 12-well plates at a density of 3 × 10⁵ cells per well 24 h before transfection, reaching approximately 60-80% confluence at the time of transfection. Transfection was performed using Lipofectamine RNAiMAX (#13778-150, Thermo Fisher Scientific) following the standard protocol. Briefly, siRNA–Lipofectamine complexes were prepared in Opti-MEM (#31985070, Thermo Fisher Scientific) using a final siRNA concentration of 12.5 pmol per well, incubated for 5 min at room temperature, and added dropwise to the cells. A Stealth scrambled siRNA (Invitrogen) was used as a negative control at the same concentration. Cells were incubated for 48 h post-transfection before harvesting for ATF4 and ULBP1 protein expression analysis by immunoblotting.

Buffy coats were collected from de-identified healthy donors through the blood center at Sahlgrenska University Hospital. Separation of peripheral blood mononuclear cells (PBMCs) was performed using density gradient centrifugation with Lymphoprep (Stemcell Technologies, #07861). Primary natural killer (pNK) cells were isolated with the Human NK Cell Isolation Kit (Miltenyi Biotec, #130-092-657). The purity of the isolated NK cells was verified by CD56/CD3 staining.

NK cells were pre-stimulated with 500 IU/mL IL-2 for three days prior to use in functional assays. KELLY cells (non-treated and treated with 1 μM Erastin) were used in assays with primary NK cells. KELLY cells were labelled with CellTrace Violet (1:2000, Invitrogen) and co-cultured with NK cells for three hours at the indicated effector-to-target (E:T) ratios. Cells were then stained with Live/Dead-Near IR dye (1:1000, Invitrogen). For degranulation assays, an anti-CD107a antibody (1:100, clone H4A3, BD Biosciences, #565113) was added before co-incubation.

NB cells (KELLY, SK-N-AS, and SK-N-FI) were plated for 48 h in 6-well plates (0.1/0.2x10⁶ cells per well) before 48 h treatment with Erastin (1 μM/1.5 μM) to induce ferroptosis. NB cells were labeled with CellTrace Violet (1:2000, Invitrogen, #C34557), co-cultured with NK-92 cells for three hours at the indicated E:T ratio, and subsequently stained with Live/Dead Near-IR dye (1:1000, Invitrogen, #L34993) and washed twice in PBS to determine the percentage of dead NB cells using flow cytometry.

SK-N-AS cells were stained for the expression of other NKG2D ligands: ULBP2/5/6 (1:100, clone 16S903, R&D, mouse anti-human, #MAB1298), ULBP3 (1:100, clone 166510, R&D, mouse anti-human, #MAB1517), ULBP4 (1:100, clone 6E6, Santa Cruz Biotechnology, mouse anti-human, #53133), MICA/B (1:20, BD Pharmingen, #558352). NK-92 cells were stained for NKG2D expression (1:20, Miltenyi Biotec, clone REA797, #130-111-645).

SK-N-AS and NK-92 cells were washed twice with PBS and stained with indicated mouse anti-human primary (ULBPs) or directly conjugated antibody (MICA/B, NKG2D) for 30 min at 4°C with L/D Near-IR dye (1:1000, Invitrogen). After two washes with PBS, cells stained with unconjugated primary antibodies incubated with goat anti-mouse secondary antibody (1:500, Invitrogen, A55747) for 30 minutes at 4°C. SK-N-AS cells then washed three times with PBS, and NKG2D ligand and receptor expression was analyzed by flow cytometry.

To stimulate ferroptosis, KELLY, SK-N-AS, SK-N-BE (2), and SK-N-FI were plated in 6-well plates (0.5x10⁶ cells per well) for 48 h and afterwards treated with indicated Erastin concentrations (KELLY:1.5 µM, SK-N-AS, SK-N-BE (2), and SK-N-FI: 3 µM). For lipid peroxidation measurements, cells (0.1x10⁶ per well) were plated in 96-well plates with the BODYPI™ 581/591 C11 probe (10 µM, Invitrogen, #D3861) and incubated for 5 h at 37°C in the dark prior to repeated PBS washing. By measuring the ratio of green to red fluorescence, it is possible to relatively quantify the extent of lipid peroxidation in live cells or tissue.

Separately, NB cells (KELLY, SK-N-AS, SK-N-BE (2), and SK-N-FI) were seeded and washed with PBS, then stained for 30 minutes at 4°C with anti-ULBP1 primary antibody (1:100, clone 170818, R&D systems, #MAB-1380-500) prior to washing and 30 min incubation with secondary antibody (1:500, Invitrogen, #A55747). After three washing steps with PBS, the MFI of ULBP1-expressing live cells was measured using flow cytometry. Data acquisition was performed using a five-laser BD LSRFortessa (BD Biosciences, NJ, USA), and data analysis was conducted using FlowJo software versions 10.9.0 and 10.10.0.

Cell line RNA-Seq: KELLY and SK-N-AS cells were seeded (0.5x10⁶) in 6-well plates in triplicate before treatment with Erastin for 24 hours. Control samples were harvested after 24 h. RNA was extracted using the ReliaPrep RNA Miniprep System (Promega #6111) according to the manufacturer’s protocol. The purified RNA was sent to Novogene UK for library preparation and sequencing. RNA-seq paired-end reads (read length 150 base pairs) were aligned to the human GRCh38 reference genome with an average alignment efficiency of 90.9%. Genes were annotated using GENCODE Release 29/GRCh38.p12 (https://www.gencodegenes.org/human/release_29.html) and quantified with HTSeq (version 2.0.3) (34). Only protein-coding genes were selected for further analysis. Differential gene expression was determined using DESeq2 (version 1.42.1) (35). Differentially expressed genes (DEGs) were identified using a threshold of p-value ≤ 0.05 and fold change ≥ 1.5. Over-representation analysis (ORA) was performed with the clusterProfiler package (version 4.10.1) (36) to identify enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Gene Ontology (GO) enrichment analysis was further conducted to predict the biological functions associated with the DEGs. Data visualization, including volcano plots and cnetplots, was carried out using R packages.

Kaplan-Meier analysis with calculation of log-rank test for overall and event-free survival probability in relation to the expression levels was performed using the Kaplan-scan cutoff method in ‘R2: Genomic Analysis and Visualization Platform’ (http://r2.amc.nl) with the publicly available datasets ‘Tumor Neuroblastoma-SEQC-498-rpm-Seqnb1’ (n=498) and the Cangelosi - 786 - custom - dgc2102 (n=786). The Kaplan-Scan cutoff method examines every increasing expression value as a cutoff for the log-rank test in order to find the optimal segregation point of two groups based on gene expression. This method then presents the most statistically significant cutoff, along with the corresponding Bonferroni-corrected P-value and the initial non-corrected P-value.

Statistical analyses were performed using either GraphPad Prism 7/8 software or the R statistical package (version 4.0). Statistical tests are indicated in the respective sections and figure captions.

Previous studies have demonstrated that cystine/glutamate antiporter metabolism is a crucial metabolic rewiring process for the progression and survival of NB (22, 23). Taking this into account, we first investigated whether the expression of the cystine/glutamate antiporter genes SLC3A2 and SLC7A11 correlates with the outcome of NB patients. The relationship between SLC3A2 and SLC7A11 expression and neuroblastoma survival data was evaluated using the Sequencing Quality Control (SEQC) cohort of 498 patients using the R2 database (http://r2.amc.nl). This revealed a statistically significant correlation between the increased expression of SLC3A2 and SLC7A11 genes and a poor overall survival (Figures 1A, B; P = 1.76e-19 and P = 1.87e-04, respectively). Similar results were obtained from the larger Cangelosi cohort (n=786), which also showed that higher expression levels of both SLC3A2 and SLC7A11 were significantly associated with poor overall survival rates in NB patients (Figures 1C, D; P = 2.45e-30 and P = 1.03e-07, respectively), suggesting that retained glutamate transportation could be associated with poor outcome.

Next, we set out to investigate whether the cystine/glutamate antiporter inhibitor Erastin affects NB cell growth and foci formation. For this, we used four NB cell lines with different genetic backgrounds: KELLY, SK-N-AS, SK-N-BE (2), and SK-N-FI. Consistent with previous findings (22), Erastin inhibited NB cell growth in a dose-dependent manner. NB cell lines KELLY, SK-N-AS, and SK-N-BE (2) showed similar sensitivity to Erastin. Notably, KELLY cells exhibited a more pronounced decrease in viability even at lower concentrations of Erastin compared to the other NB cell lines, suggesting heightened sensitivity to cystine-glutamate antiporter inhibition. In contrast, SK-N-FI exhibited moderate sensitivity to cystine/glutamate antiporter inhibition (Figure 2A). These differences in Erastin sensitivity likely reflect the distinct genetic and metabolic backgrounds of the NB cell lines. We then treated the NB cells with Erastin for a long-term period (14 days), and foci formation was inhibited at both 1 µM and 5 µM concentrations across all four cell lines. Similar to the proliferation assays, KELLY cells showed a marked reduction in foci formation even at lower Erastin concentrations, whereas SK-N-FI cells displayed only moderate sensitivity to Erastin treatment (Figure 2B).

Several studies have demonstrated that Erastin, a cystine/glutamate antiporter inhibitor, can induce ferroptosis in both in vitro and in vivo settings (22, 26, 31). To test this, we treated NB cells with either the ferroptosis inhibitor Liproxstatin-1 (Lip-1), Erastin, or a combination of both, and measured cell proliferation (Figure 3A). As seen in Figure 3A, the cell death induced by Erastin was prevented when NB cells were treated concomitantly with the ferroptosis inhibitor, Lip-1. The availability of cystine/cysteine is the rate-limiting step in GSH synthesis, which counteracts ROS and prevents ferroptosis — an iron-dependent form of regulated cell death caused by ROS-mediated lipid peroxidation (22, 26, 31). Monitoring ROS formation by flow cytometry using the lipid peroxidation sensor, C11-BODIPY, revealed a significant increase in cellular lipid peroxidation following Erastin treatment (Figure 3B). Collectively, these data indicate that the cystine/glutamate antiporter inhibitor Erastin indeed induces ferroptosis in NB cells.

To characterize the downstream gene expression response following ferroptosis induction, we performed a comprehensive RNA-Seq analysis after treatment of KELLY (MYCN amplified) and SK-N-AS (MYCN non-amplified) NB cells with Erastin for 24h (Figure 4A). Erastin treatment generated a robust, but variable, response across the two NB cell lines after 24 h. The total number of differentially expressed genes (DEG) (fold change ≥1.5 and P≤ 0.05) in KELLY and SK-N-AS 24 h after treatment with Erastin were 479 and 393, respectively (Figure 4A; Supplementary File 1).

To elucidate the molecular response following ferroptosis induction, we performed KEGG pathway and Gene Ontology (GO) enrichment analyses on Erastin-treated NB cells. KEGG pathway analysis revealed, not surprisingly, significant enrichment of ferroptosis-associated metabolic pathways, including arginine biosynthesis, one-carbon pool by folate, carbon metabolism, and most notably, the ferroptosis pathway itself (Figure 4B). These enriched pathways reflect metabolic reprogramming consistent with oxidative cell death and confirm Erastin’s efficacy in activating ferroptotic signaling in NB cells. Network-based KEGG visualization further highlighted key ferroptosis mediators such as SLC7A11, SLC3A2, and HMOX1, along with amino acid metabolism genes including SHMT2, PSAT1, and PHGDH, suggesting a tightly regulated redox stress response (Supplementary Figures 1A, B).

To further characterize transcriptional reprogramming, we performed GO term enrichment and network analysis of differentially expressed genes in both KELLY and SK-N-AS cells. The GO network revealed highly interconnected clusters associated with biological processes, including amino acid metabolic processes, and signaling receptor binding (Figure 4C). Notably, ATF4, a known transcriptional regulator of the integrated stress response, was prominently enriched within the oxidative stress cluster, supporting its involvement in regulating ferroptosis-associated genes. Among the upregulated genes, ULBP1 emerged as one of the most prominent hits within the signaling receptor binding category, with robust induction observed in both cell lines following Erastin treatment (Figures 4A–C). ULBP1 encodes a ligand for the NKG2D receptor expressed on NK cells that plays a pivotal role in immune recognition and anti-tumor cytotoxicity, positioning it as a key effector of ferroptosis-linked immunogenic remodeling.

Previous studies have demonstrated that the transcription factor ATF4 regulates ULBP1 expression in tumor-derived human cell lines (37, 38). To determine whether this regulatory axis is conserved in NB, we performed siRNA-mediated knockdown of ATF4 in Erastin-treated KELLY and SK-N-AS cells. Immunoblot analysis revealed that Erastin treatment upregulated both ATF4 and ULBP1 protein levels (Supplementary Figure 2), while silencing of ATF4 markedly reduced ULBP1 protein expression in the presence of Erastin (Supplementary Figure 2). These findings indicate that ULBP1 induction in response to ferroptotic stress is ATF4-dependent in NB cells, linking redox-driven stress signaling to immune recognition pathways.

To determine the clinical relevance of ULBP1 expression, we analyzed overall survival data from the Sequencing Quality Control (SEQC) neuroblastoma cohort (n = 498) using the R2: Genomics Analysis and Visualization Platform (http://r2.amc.nl). In the complete cohort (Figure 5A), patients with low ULBP1 expression (n = 224) exhibited significantly poorer overall survival compared to those with high expression (n = 274; P = 1.01e–09). This association remained consistent in clinically aggressive subsets (3, 39) defined by MYCN amplification (Figure 5B; P = 0.012), with progressive disease (Figure 5C; P = 3.39e–04) and, among patients with INSS stage 4 and 4s tumors (Figure 5D; P = 3.00e–03). Together, these results highlight ULBP1 as a potential prognostic biomarker in NB and support its involvement in Erastin-mediated ferroptosis and link to immune activation.

ULBP1, a member of the ULBP family, plays a pivotal role in activating natural killer (NK) cells through its interaction with the NKG2D receptor. ULBP1 is a key determinant of tumor immune visibility — while its downregulation supports immune evasion, its expression promotes NK cell-mediated recognition and cytotoxicity, thereby enhancing anti-tumor immune responses. Despite its potential as an immune-modulating factor, ULBP1 expression is frequently reduced in many cancers, thereby impairing NK cell-mediated tumor surveillance. To further understand the expression patterns of NKG2D ligands in NB, we analyzed data from the Human Protein Atlas (40–42). Our analysis revealed that the expression of all NKG2D ligands (including ULBP1-6, MICA, and MICB) was notably lower in NB compared to other cancer types (Supplementary Figures 3, 4).

To validate the involvement of ULBP1 in ferroptosis-associated immune modulation, its expression was analyzed in four NB cell lines: KELLY, SK-N-AS, SK-N-BE (2), and SK-N-FI. Cells were treated with Erastin, and ULBP1 expression was measured using flow cytometry (Figure 6A). Across all four cell lines, a significant increase in ULBP1 expression was observed following Erastin treatment, indicating a consistent upregulation of ULBP1 in cells undergoing ferroptosis (Figure 6A; Supplementary Figure 5A). To assess whether ferroptotic stress broadly affected NKG2D ligand expression, we additionally examined the surface expression of other NKG2D ligands (ULBP2–6, MICA, and MICB) in SK-N-AS cells following Erastin treatment. Consistent with our RNA-seq data, no significant changes in the expression of these ligands were observed, indicating that Erastin-induced ferroptosis selectively upregulates ULBP1 rather than globally enhancing NKG2D ligand expression (Supplementary Figure 5B; Supplementary File 1). To explore the potential link between ferroptosis-induced ULBP1 upregulation and NK cell-mediated cytotoxicity, co-culture experiments were performed using NB cells and the NK cell line NK-92. Before co-culture, surface expression of the activating receptor NKG2D on NK-92 cells was confirmed by flow cytometry, verifying their capacity to recognize NKG2D ligand–expressing target cells (Supplementary Figure 5C). KELLY, SK-N-AS, and SK-N-FI NB cell lines were pre-treated with Erastin to induce ferroptosis or co-treated with the ferroptosis inhibitor Lip-1. After treatment, NB cells were co-cultured with NK-92 cells, and cytotoxicity was assessed by flow cytometry.

Our results demonstrated that Erastin treatment significantly enhanced NK cell-mediated cytotoxicity in all three NB cell lines (Figure 6B). Interestingly, SK-N-FI cells displayed only moderate sensitivity to Erastin in proliferation and foci formation assays, but co-culture with NK cells revealed a significant increase in NK cell-mediated cytotoxicity, indicating that ferroptotic stress can potentiate immune-mediated cytotoxic responses (Figure 6B). However, when the ferroptosis inhibitor Lip-1 was added alongside Erastin, the observed enhancement in NK cell cytotoxicity was significantly reduced (Figure 6B). To further validate these findings using clinically relevant effector cells, we next performed co-culture experiments with primary NK cells isolated from healthy donors. Consistent with the results obtained using NK-92 cells, Erastin pretreatment of NB cells resulted in a significant increase in primary NK cell–mediated cytotoxicity compared to untreated controls (Supplementary Figure 5D). In addition, Erastin pretreatment significantly increased NK cell degranulation, as measured by surface expression of CD107a, indicating enhanced functional activation of primary NK cells upon engagement with ferroptosis-stressed NB cells (Supplementary Figure 5E). Collectively, these data demonstrate that Erastin-induced ferroptotic stress enhances NK cell cytotoxic function in both NK-92 cells and primary human NK cells.