Application of a novel myristoylproteomics approach identifies GLIPR2 as a key pro-ferroptotic substrate in non-small cell lung cancer

TL;DR

A new method to study myristoylated proteins reveals GLIPR2's role in promoting ferroptosis in lung cancer cells.

Contribution

A novel myristoylproteomics workflow identifies GLIPR2 as a key pro-ferroptotic substrate.

Findings

NMT1 and NMT2 expression correlates with ferroptosis sensitivity in NSCLC cells.

GLIPR2 is a novel myristoylated protein with elevated modification in ferroptosis-sensitive cells.

GLIPR2's pro-ferroptotic activity depends on its N-myristoylation modification.

Abstract

Protein myristoylation, catalyzed by N-myristoyltransferases (NMT1 and NMT2), is a key co- and post-translational modification involved in cellular signaling, yet its role in ferroptosis remains poorly defined. Here, we developed a novel myristoylproteomics workflow leveraging click chemistry to comprehensively profile N-myristoylated proteins within NSCLC cells. We found that NMT1 and NMT2 expression positively correlates with ferroptosis sensitivity. Genetic or pharmacological inhibition of NMT attenuated ferroptosis, whereas their overexpression enhanced it. Using our optimized quantitative myristoylproteomics platform, we identified GLIPR2 as a novel myristoylated protein with elevated modification levels in ferroptosis-sensitive cells. Functional studies confirmed that GLIPR2 promotes ferroptosis, and this pro-ferroptotic activity of GLIPR2 requires its N-myristoylation, as a myristoylation-deficient mutant (G2A) failed to restore ferroptosis sensitivity. Our work establishes a robust proteomic methodology for mapping protein myristoylation and uncovers the NMT1/NMT2–GLIPR2 axis as a pivotal regulator of ferroptosis. These findings propose new therapeutic strategies for NSCLC.

Image 1

•Developed and optimized a novel click chemistry-based myristoylproteomics approach for systematic identification of NMT substrates.•Discovered that NMT1/NMT2 expression positively correlates with ferroptosis sensitivity in NSCLC cells.•Screened out GLIPR2 as a new myristoylated protein (higher modification in ferroptosis-sensitive cells) via this technique.•Functional validation showed GLIPR2 exerts pro-ferroptotic effect depending on its myristoylation modification.

Developed and optimized a novel click chemistry-based myristoylproteomics approach for systematic identification of NMT substrates.

Discovered that NMT1/NMT2 expression positively correlates with ferroptosis sensitivity in NSCLC cells.

Screened out GLIPR2 as a new myristoylated protein (higher modification in ferroptosis-sensitive cells) via this technique.

Functional validation showed GLIPR2 exerts pro-ferroptotic effect depending on its myristoylation modification.

Figures

NMT1 and NMT2 expression positively correlates with sensitivity to ferroptosis. A–B. In non-hematopoietic cancer cell lines, elevated NMT1/2 expression correlates with reduced resistance to ferroptosis inducers. CTRP database analysis illustrates (A) correlations between gene expression and resistance to erastin, RSL3, and ML162, and (B) correlation of NMT1/2 expression with resistance to each compound. Values represent z-scored Pearson correlation coefficients. C. Dose-response curves of Calu-1 and H460 cells treated with erastin, RSL3 or ML162 for 24 h. Viability was assessed and normalized

NMT1 and NMT2 are required for ferroptosis sensitivity. A. NMT1 protein levels in control and NMT1-knockout Calu-1 cells. B. Dose-response curves of control and NMT1-KO Calu-1 cells to erastin or RSL3. C. NMT2 protein levels in control and NMT2-knockout cells. D. Dose-response curves of control and NMT2-KO cells. E. NMT1 and NMT2 protein expression in control, single-KO, and NMT1/2 double-knockout (DKO) cells. F. Dose-response curves of control and knockout cells to erastin or RSL3. G. Viability of H460-derived control, NMT1-OE, and NMT2-OE cells after 24 h treatment with erastin or RSL3. H. W

Augmented myristoylation is a feature and consequence of ferroptosis in sensitive cells. A. Workflow for metabolic labeling with YnMyr and detection of myristoylated proteins via CuAAC. B. Viability of Calu-1 cells pretreated with DMSO or 1 μM IMP-1088, then exposed to erastin or ML162 ± Fer-1. C. The impact of ferroptosis to N-myristoylation was detected by western blot. Calu-1 and H460 cells were first incubated with YnMyr for 18 h. For the final 8 h of YnMyr incubation, the cells were co-treated with 5 μM erastin in the presence or absence of 2 μM Fer-1. In a separate experiment, for the fi

Comparative myristoylome profiling identifies candidates associated with ferroptosis sensitivity. A. Experimental strategy to identify N-myristoylated proteins with higher levels in ferroptosis-sensitive Calu-1 vs. H460 cells. B. Volcano plot for Calu-1 cells shows YnMyr incorporation versus significance in the presence of IMP-1088. Confidence categories: high (blue), medium (green; circle, rhombus, triangle), low (yellow), potential (purple), non-substrates (gray). C. Corresponding volcano plot for H460 cells. D. Comparison of YnMyr intensities between Calu-1 and H460 cells. Subtypes: Calu-1

Myristoylation-dependent ER localization is required for GLIPR2 to promote ferroptosis. A-B. Dose-response curves of GLIPR2 knockout Calu-1 cells with or without GLIPR2-WT or G2A overexpressed, following treated with erastin (A) or ML162 (B) for 24 h. Viability was assessed and normalized to control. C. Western blot analysis validating the expression of GLIPR2 (WT or G2A) in reconstituted GLIPR2-knockout Calu-1 cells. EV, empty vector. D. Western blot analysis of GLIPR2 myristoylation in GLIPR2-knockout Calu-1 cells reconstituted as indicated, and treated with or without 1 μM IMP-1088. E. Immu

NMT1/2 enhances ferroptosis sensitivity in vivo through GLIPR2 myristoylation A. Western blot analysis of NMT1, NMT2, and GLIPR2 in Calu-1 xenograft models with or without IKE treatment following NMT1/2 overexpression or GLIPR2 knockout, with or without IKE treatment (50 mg/kg, 2 weeks). B. Effect of IKE on tumor growth and body weight in Calu-1 xenograft models with altered NMT1/2 or GLIPR2 expression. Mice bearing subcutaneous Calu-1 xenografts stably overexpressing NMT1/2 or with GLIPR2 knockout were treated with IKE or vehicle control. IKE was administered daily at a dose of 50 mg/kg for t

Full text

Introduction

1

Ferroptosis—a type of programmed cell death driven by iron-dependent lipid peroxidation—has gained significant attention in cancer research since its discovery in 2012 [1]. The core mechanism involves disruption of redox homeostasis: inactivation of glutathione peroxidase 4 (GPX4) or glutathione depletion triggers membrane lipid peroxidation [[2], [3], [4], [5]]. When intracellular antioxidants fail to clear excess reactive oxygen species (ROS), iron-catalyzed Fenton reactions amplify lipid peroxide production, ultimately inducing cell death [[6], [7], [8], [9]].

Tumor cells exhibit varying susceptibility to ferroptosis, influenced by regulatory molecules and their post-translational modifications. For instance, ferroptosis suppressor protein 1 (FSP1) requires N-terminal myristoylation for localization to the plasma membrane and lipid droplets, where it reduces coenzyme Q10 (CoQ10) using NADPH to trap radicals and inhibit ferroptosis independently of GPX4 [[10], [11], [12], [13]]. Similarly, zinc finger DHHC domain-containing protein 8 (zDHHC8) stabilizes GPX4 via palmitoylation, enhancing ferroptosis resistance [14]. Our previous work also showed that endogenous glutamate inhibits yes-associated protein (YAP), reducing transcription of ferritin heavy chain (FTH1) and increasing ferroptosis susceptibility in advanced lung adenocarcinoma (LUAD) [15]. Moreover, multiple metabolic pathways—including those involved in iron, lipid, and redox homeostasis—converge to fine-tune cellular sensitivity to ferroptosis [[16], [17], [18], [19]]. Thus, a deeper understanding of how protein modifications, such as myristoylation, regulate ferroptosis is essential. In particular, it remains unclear whether myristoylation regulates ferroptosis through novel substrates beyond its known role in FSP1.

Protein myristoylation is an essential lipid modification catalyzed by N-myristoyltransferases NMT1 and NMT2, critical for regulating protein membrane localization and stability. While the two enzymes exhibit partial substrate selectivity, they largely target overlapping pools of proteins, with most canonical N-terminal glycine-containing substrates being recognized and modified by both. This functional redundancy stems from their high sequence identity (76–77%) and structurally conserved catalytic domains, which enable cross-catalytic capability for most common substrates [20,21]. Substrate selectivity emerges only in specific contexts, as exemplified by the mitochondrial protein NADH:ubiquinone oxidoreductase complex assembly factor 4 (NDUFAF4) showing preferential regulation by NMT1, while Acyl-CoA binding domain containing 6 (ACBD6) demonstrates stronger functional association with NMT2 [22,23]. Growing evidence links myristoylation to the regulation of ferroptosis. For instance, FSP1 requires myristoylation at the glycine (Gly, G) 2 site to localize to the plasma membrane and exert anti-ferroptotic effects (reducing CoQ to eliminate lipid peroxides independently of GPX4) [10,11]. ACSL1 can enhance FSP1 myristoylation to boost its membrane translocation and stability, strengthening ferroptosis resistance in ovarian cancer [24]. Additionally, the inhibitor icFSP1 targets myristoylated FSP1 to induce ferroptosis in pancreatic cancer models [25,26]. However, whether NMT1/NMT2 directly regulate ferroptosis—and whether they act through other myristoylated proteins—remains unknown. Although ferroptosis inducers (e.g., erastin and RSL3) are available, and regulators such as FSP1 inhibitors continue to be developed [2,[27], [28], [29]], key knowledge gaps persist regarding the role of myristoylation: Does it have pro-ferroptotic functions? How does NMT activity affect ferroptosis sensitivity? Are there pro-ferroptotic myristoylated proteins? Addressing these questions could reveal new targets to overcome ferroptosis resistance.

These unresolved questions are compounded by significant technical limitations in myristoylation detection. While metabolic labeling with click-compatible analogs (e.g., YnMyr) has become a common tool, it faces critical challenges: its typically low proteomic coverage (often detecting only 30–100 proteins) and inability to pinpoint modification sites limit the discovery of novel substrates [21]. Furthermore, issues with labeling efficiency in lipogenic cells, cross-labeling of other lipidated proteins leading to false positives, and incompatibility with tissue samples restrict its application in physiologically relevant models [30,31]. These methodological constraints create a direct bottleneck, impeding the identification of new myristoylated regulators and leaving the broader functional landscape of myristoylation in ferroptosis—particularly its potential pro-ferroptotic roles—largely unexplored.

Therefore, this study aims to address these gaps using NSCLC models. We will: 1) investigate whether NMT1/NMT2 expression and activity modulate ferroptosis sensitivity; 2) identify and validate novel pro-ferroptotic myristoylated proteins; and 3) elucidate their functional mechanisms. To achieve this, we established a sensitive myristoylation detection workflow combining YnMyr metabolic labeling, CuAAC-based enrichment, and proteomic analysis, optimized for ferroptosis-sensitive cellular contexts. Our findings may offer new theoretical insights and targeting strategies for overcoming ferroptosis resistance in cancer.

Materials and methods

2

Chemical tools

2.1

CuSO_4_ (#451657), Tris(2-chloroethyl) phosphate (TCEP, #C4706), Tris (benzyltriazolylmethyl) amine (TBTA, #678937), myristic acid (#M3128), ML162 (#SML2561), RSL3 (#SML2234), ferrostatin-1 (Fer-1, #SML0583) and Dimethyl sulfoxide (DMSO, #472301) were obtained from Sigma-Aldrich (St Louis, MO, USA). Tetradec-13-ynoic acid (YnMyr, #RL-2055) was purchased from Iris Biotech GMBH (Würzburg, BY, Germany). IMP-1088 (#HY-112258), erastin (#HY-15763) were purchased from MedChemExpress (Monmouth, NJ, USA). DADPS Biotin Azide (AzDB, #CCT-1330) and TAMRA Biotin Azide (AzTB, #CCT-1048) were purchased from VectorLabs (Newark, CA, USA). AzRB was purchased from WuXi AppTec (Wuxi, Jiangsu, China), it was synthesized as previously described [32].

Cell culture

2.2

Calu-1, NCI-H460, LCLC-103H, EBC-1, ABC-1, HCC4006, NCI-H226, A-427, SK-MES-1, NCI-H1793, PC-9, A549 and HEK-293T were bought from Fuheng Biotechnology (Shanghai, China). Calu-1 was maintained in McCoy’s 5A medium (BasalMedia, Shanghai, China); NCI-H460, LCLC-103H, HCC4006, NCI-H226, NCI-H1793, PC-9, and A549 cells were cultured in RPMI-1640 (HyClone, Logan, UT, USA); EBC-1, ABC-1, A427, and SK-MES-1 cells were cultured in MEM (BasalMedia, Shanghai, China); and HEK-293T cells were cultured in high-glucose DMEM (HyClone, Logan, UT, USA). All media were supplemented with 10% fetal bovine serum (FBS, BDBIO, Hangzhou, Zhejiang, China) and 1% penicillin/streptomycin (Life-iLab, Shanghai, China). Cells were incubated at 37°C in a humidified atmosphere with 5% CO₂ and routinely tested for mycoplasma contamination.

Mice experiments

2.3

To establish cell-derived xenograft (CDX) models, Calu-1 cells (1 × 10^7^) with or without NMT1/2 overexpression and/or GLIPR2 knockout were suspended in 200 μl PBS and subcutaneously injected into athymic nude mice (BALB/c-nu, SPF [Beijing] Biotechnology, Beijing, China). Beginning 5 days after inoculation, mice were treated daily with IKE (50 mg/kg, MedChemExpress, #HY-114481) for two weeks before being euthanized.

Patient-derived xenograft (PDX) models were established using fresh human lung cancer tissues from patients with lung cancer at Shanghai Chest Hospital. The written informed consents were obtained from each patient in accordance with the guidelines of the Declaration of Helsinki. Briefly, tissues were minced into 2–3 mm^3^ fragments and subcutaneously implanted into NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice to generate the first-generation (F1) tumors. When F1 tumors reached approximately 1 cm^3^ in volume, they were harvested, fragmented, and re-implanted into secondary NSG mice to establish the second-generation (F2) models. When F2 tumors reached a volume of 100–150 mm^3^, mice were randomly assigned to treatment groups and administered IKE (50 mg/kg) via intraperitoneal injection daily for two weeks. Mice were then euthanized, and tumors were collected for subsequent analysis.

All the above experiments were performed in compliance with the institutional guidelines and were approved by the Institutional Ethics Committee of Shanghai Chest Hospital (Approval No. KS23037). Humane endpoints were strictly observed: mice were euthanized if the tumor volume exceeded 1500 mm^3^ or if the tumor burden reached 20% of the body weight.

Cell viability assessment

2.4

Cells were seeded in 96-well opaque white plates (JingAn Biological, Shanghai, China) at a density of 10,000 cells/well and allowed to adhere overnight. Following treatment with indicated compounds for 24 h, viability was assessed using the CellTiter-Glo Luminescent Assay (Promega, #G7573, Madison, WI, USA). Data were normalized to DMSO-treated controls.

Cell death and lipid peroxidation assays

2.5

Cell death was quantified by flow cytometry after SYTOX Green staining (KeyGEN, #KGE2504-500, Nanjing, Jiangsu, China). Lipid peroxidation was measured using BODIPY 581/591C11 (Invitrogen, #D3861, Carlsbad, CA, USA) following manufacturer's protocols.

Myristic acid analogue (YnMyr) tagging and cell lysis

2.6

Cells were metabolically labeled by supplementing the culture medium with 20 μM of either Myr or the alkyne-tagged analog YnMyr for 18 h. To assess NMT dependence, cells were pre-treated with IMP-1088 at specified concentrations for 30 min prior to YnMyr addition. Next, cells were lysed on ice using a buffer containing 1 × PBS, 0.2% SDS, 1% Triton X-100, and cocktail (Roche, #11873580001, BS, Switzerland). Lysates were clarified by centrifugation, and supernatants were collected for protein concentration determination.

CuAAC ligation and pull-down

2.7

The CuAAC ligation was performed on 200 μg of protein using a click chemistry mixture comprising TAMRA-Biotin Azide, CuSO_4_, TCEP, and TBTA (final concentrations: 0.1 mM, 1 mM, 1 mM, and 0.1 mM, respectively) in a total volume of 100 μL. Following a 1-h incubation at room temperature, the reaction was quenched with 10 mM EDTA. Proteins were then precipitated via methanol/chloroform/water (4:1:2), and the resulting pellet was washed, dried, and dissolved in 2% SDS/PBS. After dilution to lower the SDS concentration, the solution was subjected to pull-down with streptavidin-coated magnetic beads for 90 min at room temperature. Input and supernatant (Spnt) samples were reserved, and the beads were washed before final elution in loading buffer. All samples were denatured at 100 °C for 10 min prior to analysis.

SDS-PAGE and western blot, Immunohistochemistry assay (IHC) and Immunofluorescence (IF)

2.8

These experiments were performed using a routine procedure. The primary antibodies used for SDS-PAGE and western blot were as follows: anti-NMT1 (abcam, #ab186123; 1:1000, Cambridge, Cambs, UK), anti-NMT2 (abcam, #ab224045; 1:1000), anti-ACTB (Servicebio, #GB12001-100; 1:3000, Wuhan, Hubei, China), anti-GAPDH (Servicebio, #GB12002-100; 1:3000), anti-TAMRA (abcam, #ab171120; 1:1000), anti-PRKACA (Cell Signaling Technology, #5842S; 1:500, Danvers, MA, USA), anti-SRC (Cell Signaling Technology, #2123; 1:500), anti-CHCHD3 (abcam, #ab224565; 1:1000), anti-FSP1 (SANTA CRUZ, #sc-377120; 1:1000, Dallas, TX, USA), anti-GLIPR2 (SANTA CRUZ, #sc-398529; 1:500), anti-GLIPR2 (abcam, #ab122059; 1:250)， anti-OCC1 (Thermo Scientific, #PA5-20673; 1:500, Waltham, MA, USA), anti-RFTN1 (Proteintech, 24289-1-AP; 1:2000, Chicago, IL, USA), and anti-FLAG (Thermo Scientific, #MA1-91878; 1:1000), anti-SLC7A11 (ABclonal, A2413, 1:2000, Wuhan, China), anti-LPCAT3 (Proteintech, 67882-1-Ig, 1:1000), anti-ACSL4 (Proteintech, 22401-1-AP, 1:5000), anti-GPX4 (Proteintech, 67763-1-Ig, 1:2000), anti-FTH (ABclonal, A1144, 1:1000), anti-FTL (ABclonal, A1768, 1:1000), anti-TFRC (Proteintech, 84766-4-RR, 1:2000). The primary antibodies used for IHC were anti‑4‑HNE antibody (abcam, #48506). The primary antibodies used for IF were anti-FLAG (Thermo Scientific, #MA1-91878, 1:300), anti-GM130 (Proteintech, 82705-8-RR, 1:300), anti-ERP72 (Proteintech, 14712-1-AP, 1:300) and anti-TOMM20 (Proteintech, 11802-1-AP, 1:300). Nuclear and membrane staining were performed using DAPI (Servicebio) and Dil (Beyotime, C1991S, Haimen, Jiangsu, China), respectively.

Proteomics

2.9

Proteomic sample preparation commenced with a CuAAC reaction on 2 mg of protein, employing AzDB or AzRB in place of TAMRA-Biotin Azide. Following precipitation, the solubilized protein was subjected to affinity enrichment using NeutrAvidin agarose resin. The resin was then stringently washed and proteins underwent on-bead processing including reduction and alkylation. Digestion protocols were probe-specific: AzRB samples were digested with trypsin to generate combined samples. In contrast, AzDB samples underwent LysC/trypsin digestion, producing “protein samples” from the supernatant, while bound peptides were acid-eluted to generate distinct “site samples”. All final samples were prepared for MS analysis by desalting and vacuum concentration.

LC-MS/MS analysis

2.10

LC-MS/MS analyses for both protein and site-specific samples were perform-ed on the same instrument platform (Vanquish Neo UPLC/Orbitrap Exploris 480). Shared parameters included: a column (25 cm × 75 μm), flow rate (300 nL/min), mobile phases (A: 0.1% FA in water; B: 0.1% FA in 80% ACN), and MS settings (full scan at 60k resolution; dd-MS2 on top 21 precursors with HCD NCE 30% at 15k resolution). The gradients were optimized for each sample type: a longer, shallower gradient was used for protein samples, while a shorter, steeper gradient was applied for site-specific analysis. The specific m/z scan range was 350-1800 for protein samples and 350-2000 for site samples. The authors acknowledge ChomiX Biotech Co., Ltd. for their assistance with the LC-MS/MS work.

LC-MS/MS analysis for ferroptosis-related metabolites (PE-AA, PE-AdA, PE-AdA-OOH, and PE-AA-OOH) was performed on an AB 6500+ triple quadrupole tandem MS system equipped with an electrospray ionization (ESI) source. Chromatographic separation was achieved using a C18 column with a gradient of mobile phases A (water with 0.1% formic acid) and B (acetonitrile/isopropanol with 0.1% formic acid). The mass spectrometer operated in positive ion mode with multiple reaction monitoring (MRM). Data were acquired and processed using Analyst Software 1.7.1.

Proteomic data analysis

2.11

Raw MS data were processed with Proteome Discoverer 2.5 by searching against the UniProt Homo sapiens database. Search parameters included: cysteine carbamidomethylation (fixed modification); N-terminal acetylation and methionine oxidation (variable modifications); trypsin digestion (max two missed cleavages). Mass tolerances were set at 10 ppm for precursors and 0.02 Da for fragments. Identifications were filtered at 1% FDR at peptide, protein, and site levels. Quantification was performed using the “unique and razor peptides” mode, and peptides shorter than six amino acids were excluded. Data were processed using Microsoft Office Excel 2020 and Perseus. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the iProX partner repository with the dataset identifier IPX0015833000.

Identification of YnMyr-modified peptides with Proteome Discoverer

2.12

Myristoylated peptides were identified in Proteome Discoverer 2.5 by including a variable modification of +349.273 Da on protein N-terminal glycine residues. To ensure quantification accuracy and reduce potential artifacts from incomplete digestion or ambiguous peptide-to-protein mappings—such as peptides shared across multiple proteins—we initially performed deduplication and integration at the modification site level. Missing values were replaced with zeros to allow summation of quantitative values for the same modification site. When a peptide matched multiple proteins, entries were split according to the corresponding protein IDs. Modification sites annotated under “Modifications in Master Proteins” were merged and grouped by unique sites, and abundance values were summed to aggregate signals from multiple peptides mapping to the same site. Following data integration, a filter was applied to retain only modification sites quantified by a minimum of two valid values in the YnMyr-labeled group. Quantitative ratios were calculated as the median of all pairwise abundance ratios between experimental and control samples, with a maximum cap of 100 to minimize the influence of outliers. A one-sample t-test was used to assess whether the ratios significantly differed from 1 (indicating no change). Sites with imputed zero values that lacked sufficient valid data were excluded from ratio calculation and statistical testing to avoid bias.

Identification of YnMyr-modified peptides with PEAKS

2.13

MS data were analyzed with PEAKS 10.6 using the Homo sapiens (UniProt) database. Trypsin digestion was simulated with up to three missed cleavages permitted. The mass tolerance was defined as 10 ppm for precursor ions and 0.02 Da for product ions. For post-translational modifications, cysteine carbamidomethylation was designated as fixed, whereas methionine oxidation and a +349.273 Da addition on peptide N-terminal glycine were considered variable. Peptide identification was controlled at an FDR ≤0.01.

Transient transfection

2.14

siRNAs were obtained from GenePharma (Shanghai, China). Transfection was carried out using the EZ Trans Cell Transfection Reagent (Life-iLab, #AC04L092).

Quantitative PCR

2.15

Reverse transcription was carried out using the HiScript III RT SuperMix (Vazyme, Nanjing, Jiangsu, China) to synthesize cDNA. Quantitative PCR was conducted with ChamQ Universal SYBR qPCR Master Mix (Vazyme). Primer sequences used were as follows:

GLIPR2-F: AAGGCCCACAATGAGTACCG.

GLIPR2-R: CAAGGTTCTCCCCACACTGG.

RFTN1-F: CTGTACCAGCAGGGCTTCTC.

RFTN1-R: TGGTCTGTCGGGTGGTCTAA.

OCC1-F: ATGGAGGAGTATATGTTGGCCT.

OCC1-R: ATTACAACGTCCCACGGAGC.

FYN-F: ATGGGCTGTGTGCAATGTAAG.

FYN-R: GAAGCTGGGGTAGTGCTGAG.

OGFRL1-F: CGGTTCGCCCAGAAACACTA.

OGFRL1-R: TTTGCTTCCCTCCGACTGTTC.

ACTB-F: CATGTACGTTGCTATCCAGGC.

ACTB-R: CTCCTTAATGTCACGCACGAT.

CRISPR–Cas9-mediated gene knockout

2.16

Single guide RNAs (sgRNAs) targeting NMT1, NMT2, or GLIPR2 were inserted into the lentiCRISPR v2-puro vector to create single-gene knockout cell lines. For double knockout of NMT1 and NMT2, sgRNAs targeting NMT1 were cloned into the lentiCRISPR v2-blast vector. Lentiviral particles were generated using the Lentivirus Packaging Kit (ZORIN, #ZR-LPK-001, Shanghai, China). To obtain NMT1/NMT2 double-knockout cells, NMT2-knockout cells were infected with lentiCRISPR v2-Blast virus and selected using 6 μg/mL blasticidin (Beyotime, #ST018). The sgRNA sequences used were: NMT1 sgRNA1: GCCATTGAGCTGTTCTCAGT NMT1 sgRNA2: GGCCATTGAGCTGTTCTCAG NMT2 sgRNA1: GCGTCTATCCCGCACGTGTCC NMT2 sgRNA2: GTGATCTCTCGGATTAGCAC GLIPR2 sgRNA1: ACCTGGCTTGCGGCGGACGA GLIPR2 sgRNA2: TGAAGGCCCACAATGAGTAC.

Overexpressed cell lines generation

2.17

Plasmids expressing NMT1-FLAG and NMT2-FLAG were purchased from BioVision (Shanghai, China). All GLIPR2-related constructs (GLIPR2-3xFLAG, GLIPR2(G2A)-3xFLAG, COX8A-GLIPR2(G2A)-3xFLAG, Lyn11-GLIPR2(G2A)-3xFLAG, GLIPR2(G2A)-3xFLAG-Cb5, and GLIPR2(G2A)-3xFLAG-CENP-R_altORF) were generated by and purchased from Genomeditech. To prevent Cas9/sgRNA2-mediated cleavage of the integrated donor, the PAM site corresponding to sgRNA2 was mutated in the donor sequences for GLIPR2-3xFLAG and GLIPR2(G2A)-3xFLAG. The COX8A-GLIPR2(G2A)-3xFLAG and Lyn11-GLIPR2(G2A)-3xFLAG constructs were created by fusing the COX8A mitochondrial targeting signal or the first 11 amino acids (including the myristoylation and palmitoylation motifs) of Lyn kinase, respectively, to the N-terminus of GLIPR2(G2A)-3xFLAG. For GLIPR2(G2A)-3xFLAG-Cb5 and GLIPR2(G2A)-3xFLAG-CENP-R_altORF, amino acids 100–134 of cytochrome b5 (ER membrane anchor) or amino acids 5–14 of the CENP-R alternative open reading frame (CENP-R_altORF) were fused to the C-terminus of GLIPR2(G2A)-3xFLAG, respectively. Stable cell lines overexpressing these constructs were established by lentiviral transduction followed by selection with 2 μg/mL puromycin.

GSH/GSSG ratio and Fe2+ measurements

2.18

GSH/GSSG ratio was measured using the GSSG/GSH Quantification Kit II (#G263, DOJINDO, Kumamoto, Japan). Fe^2+^ was measured using the Kit from abcam (#ab83366, abcam).

Analysis of CTRP dataset

2.19

Gene expression correlations with resistance to erastin, RSL3, and ML162 were obtained from the CTRP v2 portal (https://ctd2-data.nci.nih.gov/Public/Broad/CTRPv2.0_2015_ctd2_ExpandedDataset/) [33]. RNA-seq TPM data were sourced from the DepMap 22Q2 Public dataset (https://doi.org/10.6084/m9.figshare.19700056.v2).

Statistical analysis

2.20

All data are expressed as mean ± standard deviation (s.d.) unless indicated otherwise. Differences between two groups were assessed using unpaired two-tailed Student's t-tests. For multiple group comparisons, one-way ANOVA with Bonferroni post-hoc correction was applied. Analyses were conducted in GraphPad Prism 8.0.

Results

3

Positive correlation between NMT1/NMT2 expression and ferroptosis sensitivity

3.1

Analysis of the CTRP database, which encompasses over 800 cancer cell lines, demonstrates a link between gene expression and drug response. Specifically, elevated expression of NMT1 and NMT2 correlated with increased sensitivity to ferroptosis inducers (erastin, RSL3, ML162), indicating that higher expression of these genes confers reduced resistance to ferroptosis (Fig. 1A and B). We examined the sensitivity of 12 NSCLC cell lines to ferroptosis induced by ML162, and found that Calu-1 was the most sensitive, whereas H460 was the most resistant. (Fig. S1A). In line with prior research [1,11], we utilized NSCLC models where Calu-1 and H460 cell lines represent high and low ferroptosis sensitivity phenotypes, respectively, a differential sensitivity confirmed by our cell viability assays (Fig. 1C). Furthermore, ferroptosis induced by erastin and RSL3 in sensitive Calu-1 cells was effectively rescued by the inhibitor Ferrostatin-1 (Fer-1) (Fig. 1D–F). Corroborating the database findings, NMT1 and NMT2 levels were significantly higher in ferroptosis-sensitive Calu-1 cells than in H460 cells (Fig. S1B-C).Fig. 1NMT1 and NMT2 expression positively correlates with sensitivity to ferroptosis. A–B. In non-hematopoietic cancer cell lines, elevated NMT1/2 expression correlates with reduced resistance to ferroptosis inducers. CTRP database analysis illustrates (A) correlations between gene expression and resistance to erastin, RSL3, and ML162, and (B) correlation of NMT1/2 expression with resistance to each compound. Values represent z-scored Pearson correlation coefficients. C. Dose-response curves of Calu-1 and H460 cells treated with erastin, RSL3 or ML162 for 24 h. Viability was assessed and normalized to control. D. Viability of cells treated with 5 μM erastin or 1 μM RSL3, with or without 2 μM Fer-1, for 24 h. E. Cell death was evaluated by Sytox Green staining and flow cytometry after treatment as in (D). F. Lipid peroxidation was measured by BODIPY C11 staining and flow cytometry following 8 h treatment as indicated. Data represent mean ± SD; n = 3 biologically independent samples (C–F). One-way ANOVA was used for statistical comparisons in D–F. ∗∗∗∗p < 0.0001; NS, not significant.Fig. 1

NMT1/NMT2 loss attenuates ferroptosis sensitivity

3.2

To elucidate the functional roles of NMT1 and NMT2 in ferroptosis, we created knockout clones for each gene in the sensitive Calu-1 cell line. Genetic ablation of either NMT1 or NMT2 was found to substantially diminish cellular sensitivity to ferroptosis inducers (Fig. 2A–D). Furthermore, double knockout of both NMT1 and NMT2 suppressed ferroptosis sensitivity more profoundly than single knockouts (Fig. 2E and F).Fig. 2NMT1 and NMT2 are required for ferroptosis sensitivity. A. NMT1 protein levels in control and NMT1-knockout Calu-1 cells. B. Dose-response curves of control and NMT1-KO Calu-1 cells to erastin or RSL3. C. NMT2 protein levels in control and NMT2-knockout cells. D. Dose-response curves of control and NMT2-KO cells. E. NMT1 and NMT2 protein expression in control, single-KO, and NMT1/2 double-knockout (DKO) cells. F. Dose-response curves of control and knockout cells to erastin or RSL3. G. Viability of H460-derived control, NMT1-OE, and NMT2-OE cells after 24 h treatment with erastin or RSL3. H**.** Western blot analysis of NMT1 and NMT2 in overexpressing cells. Data are shown as mean ± SD; n = 3 independent samples (A–F). Experiments in B, D, and F were repeated in triplicate. Statistical analysis by one-way ANOVA (A): ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.Fig. 2

Overexpression of NMT1/NMT2 enhances ferroptosis sensitivity

3.3

Overexpression of either NMT1 or NMT2 in H460 cells enhanced their sensitivity to ferroptosis inducers (Fig. 2G–H).

NMT enzymatic activity enhances susceptibility to ferroptosis

3.4

In addition to their role as scaffolding proteins, the primary function of NMT1 and NMT2 is to catalyze protein myristoylation. To systematically identify myristoylated proteins related to ferroptosis, we established a myristoylation detection workflow: cells were incubated with an alkyne analogue of myristic acid YnMyr or Myr (negative control). Intracellular proteins are labeled with YnMyr via NMT activity. Subsequently, copper-catalyzed azide-alkyne cycloaddition (CuAAC) was used to conjugate the metabolically labeled proteins to the capture reagent AzTB (TAMRA Biotin Azide). Global myristoylation levels were detected via TAMRA fluorescence, and specific myristoylated proteins were detected via biotin-based enrichment (Fig. 3A). To optimize the YnMyr metabolic labeling conditions, we performed a time-course and a concentration-series experiment in Calu-1 cells. We found that treatment with 20 μM YnMyr for 18 hours yielded robust myristoylation signals, as detected by streptavidin-HRP or TAMRA analysis (Fig. S2A-B). Treatment with the NMTs inhibitor IMP-1088 reduced ferroptosis sensitivity (Fig. 3B), suggesting that NMTs enzymatic activity enhances susceptibility to ferroptosis. The results showed that IMP-1088 effectively inhibited NMTs activity, as evidenced by reduced global myristoylation levels and weakened labeling of known myristoylated proteins such as SRC, PRKACA, and CHCHD3 (Fig. S3A). Notably, in ferroptosis-sensitive Calu-1 cells, stimulation with erastin or RSL3 increased global myristoylation levels, an effect not rescued by Fer-1. This phenomenon was absent in ferroptosis-resistant H460 cells (Fig. 3C). Under steady-state conditions, H460 cells exhibited lower basal myristoylation levels than Calu-1 cells (Fig. 3D). Treatment with erastin or RSL3 did not influence the protein expression of NMT1 or NMT2 (Fig. 3E-F). These results suggest that elevated myristoylation levels in Calu-1 cells may contribute to their heightened ferroptosis sensitivity.Fig. 3. Augmented myristoylation is a feature and consequence of ferroptosis in sensitive cells. A. Workflow for metabolic labeling with YnMyr and detection of myristoylated proteins via CuAAC. B. Viability of Calu-1 cells pretreated with DMSO or 1 μM IMP-1088, then exposed to erastin or ML162 ± Fer-1. C. The impact of ferroptosis to N-myristoylation was detected by western blot. Calu-1 and H460 cells were first incubated with YnMyr for 18 h. For the final 8 h of YnMyr incubation, the cells were co-treated with 5 μM erastin in the presence or absence of 2 μM Fer-1. In a separate experiment, for the final 2 h of YnMyr incubation, the cells were co-treated with 1 μM RSL3 with or without 2 μM Fer-1.D. Samples were analyzed by TAMRA (top) or enriched by pull-down on streptavidin beads and analyzed by Western blot (bottom). The sample before pull-down (Input), pull-down sample (PD) and the supernatant from the pull-down (Spnt) were analyzed. C-SRC, PRKACA and CHCHD3 were enriched in the pull-down samples. GAPDH: loading control.E-F. The impact of ferroptosis to NMT1 or NMT2 was detected by western blot. Cells (Calu-1, E; H460, F) were treated with 5 μM erastin or 1 μM RSL3 for indicated time.Experiments in B were repeated in triplicate.Fig. 3

Optimization of YnMyr and IMP-1088 labeling conditions

3.5

Furthermore, we titrated the NMTs inhibitor IMP-1088 to determine its potency in both Calu-1 and H460 cell lines. Titration of the NMT inhibitor IMP-1088 in both Calu-1 and H460 cell lines demonstrated that the inhibitor suppressed cellular NMT activity in a dose-dependent manner. A low dose of 10 nM elicited a modest reduction in myristoylation, whereas 100 nM produced a potent and significant suppression (Fig. S3B-C). Meanwhile, NMT1 and NMT2 protein levels remained unaffected by IMP-1088 treatment (Fig. S3D).

Myristoylome analysis based on novel capture technology

3.6

Given the high background signal attributed to non-myristoylation-dependent incorporation in metabolic labeling with YnMyr, direct MS-based detection of myristoylated peptides is particularly critical [[34], [35], [36], [37]]. To maximize direct MS/MS evidence for myristoylated peptides, we started with a small-scale pilot experiment to test capture reagents. Although AzRB has been validated as a tool for identifying myristoylated proteins and peptides [35], AzDB (DADPS Biotin Azide)—which is expected to enhance the discovery of myristoylated peptides—has not previously been applied to the study of protein myristoylation (Fig. S4A). We therefore evaluated AzDB for its ability to enrich YnMyr-labeled proteins and observed robust enrichment of putative NMT substrates (Fig. S4B). The pre-enrichment process was the same as the myristoylation Western blot detection workflow (Fig. 3A). For the AzRB reagent, LysC and trypsin digestion release both myristoylated peptides (sites) and unmodified peptides (other peptides from myristoylated proteins or background peptides) simultaneously for mixed detection. For the AzDB reagent, LysC and trypsin digestion first release unmodified peptides. Subsequently, myristoylated peptides remaining on the agarose resin are released using formic acid, allowing the two peptide fractions to be detected separately. This reduces sample complexity and improves the detection rate of modification sites (Fig. S4C). As expected, AzDB detected a greater number of sites compared to AzRB in H460 cells (Fig. S4D), leading us to select it for subsequent myristoylome analysis.

Proteomic screening for ferroptosis-associated myristoylated proteins Myristoylome profiling was utilized to screen for candidate proteins linked to ferroptosis sensitivity. We focused on proteins whose myristoylation was enriched in ferroptosis-sensitive Calu-1 cells over H460 cells, suggesting a potential functional association. The experimental design included Calu-1 and H460 cells subjected to four treatment groups (Myr, YnMyr, YnMyr + 10 nM IMP-1088, YnMyr + 100 nM IMP-1088), each with fourbiological replicates. All samples were ligated to AzDB and affinity enriched as described above (Fig. 3A). Following digestion, all peptides were subjected to LC-MS/MS analysis and quantified by Label Free Quantifition (LFQ, Fig. 4A).Fig. 4. Comparative myristoylome profiling identifies candidates associated with ferroptosis sensitivity. A. Experimental strategy to identify N-myristoylated proteins with higher levels in ferroptosis-sensitive Calu-1 vs. H460 cells. B. Volcano plot for Calu-1 cells shows YnMyr incorporation versus significance in the presence of IMP-1088. Confidence categories: high (blue), medium (green; circle, rhombus, triangle), low (yellow), potential (purple), non-substrates (gray). C. Corresponding volcano plot for H460 cells. D. Comparison of YnMyr intensities between Calu-1 and H460 cells. Subtypes: Calu-1 > H460 (red), unique to Calu-1 (yellow), unique to H460 (green), H460 > Calu-1 (blue), common (purple), non-substrates (gray). E. Schematic for selecting myristoylated substrates linked to ferroptosis sensitivity. F. Viability of Calu-1 cells after siRNA knockdown of candidate genes and treatment with 0.04 μM ML162. Data are mean ± SD; n = 3 (F). Statistics: unpaired t-test (F). ∗p < 0.05, ∗∗p < 0.01, NS, not significant.Fig. 4

To generate a comprehensive list of myristoylated proteins in Calu-1 and H460 cells, we integrated multiple complementary strategies: YnMyr-based metabolic labeling and enrichment, assessment of substrate responsiveness to NMT inhibition, direct identification of myristoylated peptides, and cross-referencing with previously published evidence (Multimedia component 2). Substrates were further categorized into four groups based on their confidence levels.

High confidence

3.7

Core criteria: Satisfies either of the two conditions below, combined with a dose response to NMT inhibition and significant in YnMyr/100 nM IMP-1088.

• 1.YnMyr enrichment: log_2_FC (YnMyr/Myr) > 2.
• 2.Detection of YnMyr peptide.

Medium Confidence was classified as three subcategories.

• 1.Significant in YnMyr/Myr enrichment (0.6 < log_2_FC < 2) and significant in YnMyr/100 nM IMP-1088 with dose-dependently sensitive to NMT inhibition.
• 2.YnMyr/Myr enrichment (log_2_FC > 2) or detection of YnMyr peptide with significant in YnMyr/100 nM IMP-1088.
• 3.NaN values in 100 nM IMP-1088 treatment and detection of YnMyr peptide.
• Low confidence: Detection of YnMyr peptide and/or YnMyr/Myr enrichment.

“Possible substrates” are additional candidates supported by literature documentation of myristoylation (not classified into the three confidence tiers above).

Throughout this classification, a result was considered significant if it passed a T-test with an adjusted q-value <0.05, applied to both the dose–response comparisons and the enrichment comparisons.

Based on the results, all detected myristoylated proteins were classified into four confidence levels (high-confidence [blue], medium-confidence [green], low-confidence [yellow], potential [purple]) using criteria including YnMyr/Myr enrichment ratio, IMP-1088 inhibition, presence of a YnMyr-modified peptides, and literature support (Fig. 4B-C). Volcano plots were generated for within-cell line comparisons (Calu-1: Fig. S5A-C; H460: Fig. S5D-F, Multimedia component 2), showing differential responses of myristoylated proteins to the inhibitor between the cell lines. As expected, most substrates exhibited a dose-dependent response to IMP-1088 treatment (Fig. S5C and S5F). Of the proteins identified as candidate substrates in our study, 122 out of 152 in Calu-1 cells and 120 out of 164 in H460 cells have been previously reported as myristoylated, strongly supporting the validity of our identification strategy (Multimedia component 2) [34,35,38]. Comparison of YnMyr intensities between the two cell lines revealed cell type-specific differences in myristoylation(Fig. 4D, Multimedia component 3). Analysis using both PD and PEAKs software identified distinct sets of myristoylation sites (Fig. S5G-H, Multimedia component 2). Our myristoylome profiling revealed 42 substrates with elevated myristoylation levels in Calu-1 versus H460 cells (Multimedia component 3). Among these, 29 were classified as high-confidence hits specific to the Calu-1 line. Correlation analysis using the CTRP database revealed that among these 42 proteins, only expression of five proteins, including Overexpressed in colon carcinoma 1 protein (OCC1), Raftlin lipid raft linker 1 (RFTN1), Glioma pathogenesis-related protein 2 (GLIPR2), Opioid growth factor receptor like 1 (OGFRL1), and FYN proto-oncogene (FYN), correlated with sensitivity to ferroptosis inducers(erastin, RSL3, ML162) (Fig. 4E). To identify key effectors regulating ferroptosis from the candidate myristoylated proteins, we conducted a systematic functional screen. First, we individually knocked down candidate genes using siRNA(Fig. S5I) and assessed the impact on ferroptosis sensitivity via cell viability assays. Knockdown of GLIPR2, RFTN1, and OCC1 significantly decreased cellular sensitivity to ferroptosis inducers (Fig. 4F), suggesting these proteins play positive regulatory roles in ferroptosis. Analysis of drug sensitivity data from the CTRP database further supported this, showing that high expression of GLIPR2, RFTN1, and OCC1 correlated with increased sensitivity to ferroptosis inducers such as erastin, RSL3, and ML162. Notably, GLIPR2 showed the most significant correlation (lowest Z-score), suggesting it might be a core regulator (Fig. S5J). We next verified the myristoylation status of these candidate proteins by Western blot. IMP-1088 treatment significantly inhibited the myristoylation of GLIPR2 and RFTN1 (Note: OCC1 was not tested due to lack of a suitable antibody) (Fig. S3A). Furthermore, under basal conditions, the myristoylation levels of GLIPR2 and RFTN1 were higher in Calu-1 cells than in H460 cells, consistent with their higher ferroptosis sensitivity phenotype (Fig. S5K).

GLIPR2 promotes ferroptosis via myristoylation-dependent endoplasmic reticulum localization

3.8

We generated GLIPR2 knockout Calu-1 cells and found that GLIPR2 knockout significantly reduced cellular sensitivity to ferroptosis inducers (Fig. S6A-C). Re-introduction of wild-type (WT) GLIPR2 restored ferroptosis sensitivity in GLIPR2-KO cells, whereas a myristoylation-deficient mutant (glycine 2 to alanine, G2A) [39] failed to do so, indicating that GLIPR2 myristoylation is essential for promoting ferroptosis (Fig. 5A-C). Applying AzDB to the proteomics, we directly identified the N-terminally YnMyr-modified peptides of GLIPR2 by MS/MS (Fig. S6D). We also demonstrated that both IMP-1088 treatment and the G2A mutation abolished GLIPR2 myristoylation (Fig. 5D). N-Glycine myristoylation is critical for directing proteins to specific subcellular locations by facilitating protein–protein and protein–membrane interactions [40]. Then immunofluorescence analysis clearly indicated that GLIPR2 co‑localized with the endoplasmic reticulum (ER), but not with the Golgi apparatus, mitochondria (Mito), or plasma membrane (PM). This ER co‑localization was lost for the GLIPR2 (G2A) mutant and upon IMP-1088 treatment (Fig. 5E and S6E). An ER‑targeted fusion protein, GLIPR2 (G2A)-Cb5, restored the ability to promote erastin- and ML162-induced ferroptosis, similar to wild-type GLIPR2. In contrast, GLIPR2 mutants forcibly localized to the Golgi apparatus (CENP-R altROF), the plasma membrane (Lyn11), or the mitochondria (COX8A) failed to do so [11,41] (Fig. 5F-I). Cell viability assays demonstrated that overexpression of wild‑type GLIPR2 in H460 cells enhanced erastin‑ and ML162‑induced ferroptosis, whereas the GLIPR2 (G2A) mutant lacked this effect (Fig. S6F-G). Moreover, GLIPR2 knockout attenuated the pro‑ferroptotic effect caused by NMT1/2 overexpression, suggesting that NMTs promote ferroptosis through a GLIPR2-mediated mechanism (Fig. S6H-I). Myristoylated GLIPR2 levels decreased upon NMT1/2 single knockout, with a more pronounced reduction observed after double knockout (Fig. S6J). Additional assays measuring cell death and lipid ROS confirmed that GLIPR2 knockout inhibited ferroptosis, whereas GLIPR2 overexpression potentiated it (Fig. 6A–B). Consistently, the G2A mutation abolished GLIPR2’s pro‑ferroptotic function, while the ER‑targeted GLIPR2(G2A)-Cb5 mutant exhibited activity comparable to wild-type GLIPR2 (Fig. 6C–D). Together, these results demonstrate that myristoylation-dependent ER localization is required for GLIPR2 to promote ferroptosis.Fig. 5. Myristoylation-dependent ER localization is required for GLIPR2 to promote ferroptosis. A-B. Dose-response curves of GLIPR2 knockout Calu-1 cells with or without GLIPR2-WT or G2A overexpressed, following treated with erastin (A) or ML162 (B) for 24 h. Viability was assessed and normalized to control. C. Western blot analysis validating the expression of GLIPR2 (WT or G2A) in reconstituted GLIPR2-knockout Calu-1 cells. EV, empty vector. D. Western blot analysis of GLIPR2 myristoylation in GLIPR2-knockout Calu-1 cells reconstituted as indicated, and treated with or without 1 μM IMP-1088. E. Immunofluorescence analysis showing the subcellular localization of re-introduced wild-type GLIPR2 and the G2A mutant in GLIPR2-knockout Calu-1 cells. Scale bar, 20 μm. F-G. Dose-response curves of GLIPR2-knockout Calu-1 cells reconstituted with GLIPR2 (WT) or the indicated subcellular localization mutants, treated with erastin (F) or ML162 (G) for 24 h. H. Immunofluorescence analysis of the subcellular localization of the indicated GLIPR2 mutants in GLIPR2-knockout Calu-1 cells. Scale bar, 20 μm. I. Western blot analysis of the expression levels of GLIPR2 (WT) and the indicated mutants in GLIPR2-knockout Calu-1 cells. Data are mean ± SD; n = 3 (A–B and F-G).Fig. 5. Fig. 6GLIPR2 promotes ferroptosis through myristoylation-dependent regulation of LPCAT3 and lipid peroxidation (A-B) Cell death (24 h, A) and lipid ROS (8 h, B) were measured in Calu-1 cells with or without GLIPR2 knockout, and in H460 cells overexpressing GLIPR2 (WT) or GLIPR2 (G2A). Cells were treated with erastin (5 μM) or ML162 (0.04 μM), in the presence or absence of Fer-1 (2 μM) for 24h.(C-D) Cell death (24 h, C) and lipid ROS (8 h, D) were measured in Calu-1 cells with GLIPR2, following overexpressing GLIPR2 (WT), GLIPR2 (G2A) or GLIPR2 (G2A)-Cb5. Treatments were as in (A-B).(E) Western blot analysis of indicated ferroptosis-related proteins in parental (WT) and GLIPR2‑knockout Calu-1 cells.(F-K) Levels of the lipid peroxidation products PE-AA-OOH (F) and PE-AdA-OOH (G), their precursors PE-AA (H) and PE-AdA (I), intracellular Fe^2+^ level (J) and the GSH/GSSG ratio (K) in parental (WT) and GLIPR2‑knockout Calu-1 cells treated with erastin (5 μM) or ML162 (0.04 μM) for 24 h.Data are mean ± SD; n = 3 (A-D and F-K). Statistics: one-way ANOVA (A-D) or unpaired t-test (F-K). ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001; NS, not significant.Fig. 6

GLIPR2 induces ferroptosis by upregulating LPCAT3 to drive lipid peroxidation

3.9

Next, we investigated the mechanism by which GLIPR2 regulates ferroptosis. First, we performed GLIPR2 knockout and screened a panel of potential ferroptosis‑related targets (Fig. 6E). The results showed that only LPCAT3 was significantly downregulated. Upon ferroptosis induction, GLIPR2 knockout suppressed the generation of the canonical pro‑ferroptotic lipid peroxidation products PE‑AA‑OOH and PE‑AdA‑OOH (Fig. 6F-G), and also attenuated the consumption of their precursors, PE‑AA and PE‑AdA (Fig. 6H-I). In contrast, alterations in Fe²⁺ and glutathione levels during ferroptosis remained unaffected by GLIPR2 status (Fig. 6J-K). These data suggest that GLIPR2 promotes ferroptosis specifically by modulating LPCAT3-dependent lipid peroxidation, without affecting iron accumulation or glutathione depletion.

NMT1/2 enhances ferroptosis sensitivity in vivo through GLIPR2 myristoylation

3.10

To further investigate whether NMT1/2 enhances ferroptosis sensitivity in lung cancer cells through myristoylation of GLIPR2, we conducted in vivo experiments. First, we established Calu-1 cell lines stably overexpressing NMT1/2 with or without GLIPR2 knockout (Fig. 7A) and subcutaneously inoculated them into nude mice to evaluate the effect of the ferroptosis inducer IKE on xenograft tumor growth. The results showed that IKE treatment significantly inhibited tumor growth, and this effect was more pronounced in the NMT1/2-overexpressing group (Fig. 7B-D). However, when GLIPR2 was simultaneously knocked out, the growth-inhibitory advantage conferred by NMT1/2 overexpression was largely abolished (Fig. 7B-D), indicating that the promotion of ferroptosis sensitivity by NMT1/2 depends on GLIPR2. Further analysis revealed that during IKE treatment, mice bearing NMT1/2-overexpressing tumors exhibited less body weight loss compared with those bearing wild-type tumors, suggesting a potentially lower tumor burden and better treatment response. This effect disappeared upon GLIPR2 knockout, further supporting GLIPR2 as a key downstream effector of NMT1/2 function (Fig. 7B). Meanwhile, we assessed lipid peroxidation levels in tumor tissues. Under IKE treatment, NMT1/2 overexpression significantly increased MDA content, and this elevation was reversed by GLIPR2 knockout. Collectively, these results suggest that NMT1/2 likely enhances cellular sensitivity to ferroptosis by promoting the myristoylation modification of GLIPR2. To validate the proposed mechanism in a clinically relevant model, we selected three human lung cancer tissue samples with high NMT1/2 expression and three with low NMT1/2 expression (Fig. 7F) and successfully established patient-derived xenograft (PDX) models. Following IKE treatment, immunohistochemical analysis revealed that the levels of the lipid peroxidation end product 4-HNE were significantly higher in PDX tumors from the high NMT1/2 expression group than in those from the low expression group (Fig. 7G and H), indicating that tumors with elevated NMT1/2 expression are more sensitive to ferroptosis induction. Further analysis showed that under IKE treatment, compared with the low NMT1/2 expression group, the high expression group exhibited markedly reduced cell viability, increased cell death, and significantly elevated lipid ROS levels (Fig. 7I–K). Collectively, these results from human tissue-derived models further demonstrate that NMT1/2 enhances the sensitivity of lung cancer cells to ferroptosis by regulating the myristoylation of GLIPR2.Fig. 7NMT1/2 enhances ferroptosis sensitivity in vivo through GLIPR2 myristoylation A. Western blot analysis of NMT1, NMT2, and GLIPR2 in Calu-1 xenograft models with or without IKE treatment following NMT1/2 overexpression or GLIPR2 knockout, with or without IKE treatment (50 mg/kg, 2 weeks). B. Effect of IKE on tumor growth and body weight in Calu-1 xenograft models with altered NMT1/2 or GLIPR2 expression. Mice bearing subcutaneous Calu-1 xenografts stably overexpressing NMT1/2 or with GLIPR2 knockout were treated with IKE or vehicle control. IKE was administered daily at a dose of 50 mg/kg for two weeks. Tumor volume and body weight were monitored throughout the treatment period. Data were presented as mean ± SD. n = 5; ∗∗, p < 0.01; ∗∗∗, p < 0.001. C-D. Tumor volume in Calu-1 xenograft models after IKE treatment. Athymic nude mice bearing Calu-1 xenografts with stable NMT1/2 overexpression or GLIPR2 knockout were treated daily with IKE (50 mg/kg) or vehicle control for two weeks. Tumor volume was measured and shown as a bar chart (mean ± SD, n = 5). Statistical significance is indicated: ∗∗, p < 0.01; ∗∗∗, p < 0.001, ∗∗∗∗, p < 0.0001; ns, not significant. E. Lipid peroxidation was assessed by measuring MDA levels in tumor tissues from athymic nude mice bearing Calu-1 xenografts with stable NMT1/2 overexpression or GLIPR2 knockout, after treatment with IKE or vehicle control (50 mg/kg daily for 2 weeks). Data were presented as mean ± SD. n = 5; ∗∗∗∗, p < 0.0001; ns, not significant. F. Western blot analysis of NMT1, NMT2, and GLIPR2 in patient-derived xenograft. G-H. Representative IHC staining of 4-HNE in patient-derived xenograft with the treatment of IKE or vehicle control (50 mg/kg daily for 2 weeks). IHC score was measured and shown as a bar chart (mean ± SD, n = 3). Scale bars, 100 μm ∗, p < 0.05. I-K. Cell viability, cell death, and lipid ROS generation were measured in patient-derived xenograft with the treatment of IKE or vehicle control (50 mg/kg daily for 2 weeks). n = 3; ∗∗∗∗, p < 0.0001.; NS, not significant.Fig. 7

Discussion

4

This research elucidates a novel mechanism through which NMT1 and NMT2 enhance ferroptosis sensitivity in NSCLC cells: by mediating the myristoylation of critical effector proteins. Notably, this discovery was made possible by our novel, optimized click chemistry-based myristoylproteomics workflow—a core technical advancement that overcomes longstanding limitations in detecting myristoylated proteins, thereby enabling the identification of the pro-ferroptotic substrate GLIPR2 and the validation of the NMT1/NMT2–myristoylation–GLIPR2 axis. Our findings not only expand the molecular mechanisms of the ferroptosis regulatory network but also establish this workflow as a robust tool for myristoylome profiling, offering a new theoretical foundation and technical platform for enhancing ferroptosis-based therapies by targeting the myristoylation pathway.

Analysis of the CTRP database, later validated in multiple cell lines, initially established a positive correlation between NMT1/NMT2 expression and sensitivity to ferroptosis inducers (e.g., erastin, RSL3, ML162). Notably, NMT1/NMT2 expression was significantly higher in ferroptosis-sensitive Calu-1 cells compared to resistant H460 cells. Further functional experiments demonstrated that knocking out NMT1 or NMT2, or using the myristoylation inhibitor IMP-1088, significantly reduced ferroptosis sensitivity, while overexpression of NMT1 or NMT2 enhanced sensitivity, indicating that NMT1/NMT2 act as positive regulators of ferroptosis. It is noteworthy that this study reveals the pro-ferroptotic role of NMT1/NMT2 from the perspective of “global protein myristoylation levels,” which stands in stark contrast to most previous studies focusing on the myristoylation of individual proteins (e.g., FSP1) and emphasizing their anti-ferroptotic functions [10,11], our YnMyr metabolic labeling + AzDB capture system enables large-scale, quantitative profiling of myristoylated proteins, revealing the collective contribution of NMT-mediated myristoylation to ferroptosis sensitivity. This stands in stark contrast to most previous studies focusing on the myristoylation of single anti-ferroptotic proteins and highlights the workflow's unique advantage in capturing global myristoylation dynamics. Additionally, a recent study by Gerovska et al. indicated that NMT inhibition can induce mitochondrial iron overload and parthanatos, leading to cell death in specific genetic backgrounds (e.g., KRAS/LKB1/KEAP1 co-mutated lung cancers) [42]. Although that study did not directly focus on ferroptosis, its results—combined with our findings—suggest that NMT's role in cell death regulation is context-dependent. Importantly, our workflow provides a means to dissect such context-dependent effects by quantifying how myristoylation of distinct substrates (e.g., pro-vs. anti-ferroptotic proteins) shifts across genetic or environmental conditions.

A key advantage of this study lies in the establishment of an optimized myristoylation detection workflow, which effectively addresses the limitations of existing methods and provides robust technical support for identifying ferroptosis-related myristoylated proteins. Compared to traditional radioactive labeling [43], our click chemistry-based approach (using YnMyr and AzDB) avoids health risks, shortens detection time, and enables both global myristoylation level detection (via TAMRA fluorescence) and specific protein enrichment (via biotin), significantly improving detection sensitivity and specificity. Relative to conventional click chemistry workflows that employ reagents like AzRB [21,32], our method offers a critical advancement. While AzRB enables proteome-wide profiling, it does not facilitate the separate elution of modified and unmodified peptides, thereby limiting the depth of site-specific identification. In contrast, our adoption of the AzDB reagent addresses this limitation: after enriching YnMyr-tagged proteins, LysC/trypsin digestion first releases unmodified peptides, and myristoylated peptides are then specifically acid-eluted. This separate detection reduces sample complexity and boosts modification site identification—consistent with our observation that AzDADPSB detected more myristoylation sites than AzRB.

Furthermore, our workflow optimizes labeling conditions (20 μM YnMyr, 18 h) to enhance signal intensity and reduces false positives by using Myr as a negative control and IMP-1088 to verify NMT dependence. Recent studies have also reported similar “enzyme digestion-coupled site elution” approaches, which supports the rational design of our AzDB-based strategy [44]. Furthermore, our workflow exhibits clear advantages over other enrichment strategies. For example, compared to liquid-liquid extraction methods [45], which require multiple parallel digestions (trypsin/LysC, GluC, chymotrypsin) and extractions (heptanol, octanol) with complex operations, our approach integrates negative selection of N-terminomes with solid-phase extraction, simplifying procedures while ensuring enrichment efficiency. Relative to membrane subfractionation combined with LC-MS [46], which starts from whole proteome data and faces challenges in detecting low-abundance peptides, our pre-enrichment of myristoylated proteins via click chemistry reduces background interference and enhances the detection of low-abundance substrates (e.g., GLIPR2).

Limitations and future directions

5

Although this study establishes the central role of the “NMT1/NMT2–myristoylation–GLIPR2” axis in NSCLC ferroptosis—made possible by our optimized workflow—two major limitations remain, of which can be addressed using the same technical platform:

• (1)Although our data suggest that GLIPR2 promotes ferroptosis by upregulating LPCAT3 to drive lipid peroxidation, the precise molecular link between GLIPR2 and LPCAT3 remains to be defined. It is unclear whether GLIPR2 directly regulates LPCAT3 activity, stabilizes the LPCAT3 protein, or modulates its transcription. Notably, LPCAT3 is an endoplasmic reticulum (ER)-resident enzyme that incorporates polyunsaturated fatty acids into phospholipids, a key step in ferroptotic lipid peroxidation [47]. Given that both GLIPR2 and LPCAT3 localize to the ER, their spatial proximity could facilitate functional interaction—for example, through direct binding or co-regulation within ER subdomains. Future studies should investigate whether GLIPR2 physically associates with LPCAT3, affects its enzymatic activity, or collaborates with other ER-resident factors (e.g., lipid transporters or peroxidation enzymes) to amplify pro-ferroptotic signaling.
• (2)Role of other NMT substrates unexplored: This study focused only on six core substrates. Whether NMT1/NMT2 regulate other myristoylated proteins involved in ferroptosis requires systematic validation.

From a clinical translation perspective, the findings of this study—supported by our technical workflow—have several potential applications:

• (1)Biomarker development: NMT1/NMT2 expression levels could serve as biomarkers for ferroptosis-sensitive NSCLC subtypes. Subsequent validation in clinical sample cohorts (e.g., tumor tissues from over 100 NSCLC patients) is needed to confirm their correlation with patient response to ferroptosis inducer therapies.
• (2)Therapeutic strategy optimization: For NSCLC subtypes with high NMT expression, combination therapies of “ferroptosis inducers + NMT inhibitors” could be developed. However, since NMT is widely expressed in normal tissues (e.g., liver, kidneys) [48], systemic inhibition may lead to off-target effects (e.g., impaired protein myristoylation in normal cells). Therefore, future efforts should focus on developing tissue-targeted drug delivery systems (e.g., NSCLC-specific antibody-conjugated IMP-1088 nanoparticles) to reduce toxicity.
• (3)Addressing individual differences: Ferroptosis sensitivity exhibits significant variations across cell types, tissue origins, and even gender and racial backgrounds [[49], [50], [51]]. A study by Liang et al. (2023) published in Cell revealed that ferroptosis surveillance is regulated in a sex hormone-dependent manner (e.g., estrogen can enhance ferroptosis resistance by activating the NRF2 pathway) [52]. This suggests that clinical applications must consider individual differences such as patient gender and hormone levels, potentially necessitating personalized combination therapies for different patient subgroups. Additionally, developing highly selective, tissue-targeted drug delivery systems or subcellular compartment-specific intervention strategies will be important future directions. For specific molecular subtypes of NSCLC (e.g., KRAS mutant, KEAP1 mutant), future research should optimize combination therapies (e.g., ferroptosis inducers combined with NMT targeting or immunotherapy) in more clinically relevant models (e.g., organoids, PDX models) and systematically evaluate their efficacy and safety [[53], [54], [55]].

In summary, this study not only establishes for the first time the central role of the “NMT1/NMT2–myristoylation–GLIPR2” axis in regulating ferroptosis in NSCLC but also introduces a novel, optimized myristoylproteomics workflow that addresses long-standing technical limitations in the field. This workflow—with its high sensitivity, specificity, and adaptability—serves as a powerful tool for future investigations into myristoylation's role in cancer and other diseases. Future research should integrate multi-omics data (e.g., single-cell transcriptomics, proteomics) with our myristoylation detection workflow and advanced technologies (e.g., live-cell imaging) to elucidate the dynamic regulatory mechanisms of this axis and its interactions with other pathways (e.g., iron metabolism, lipid metabolism) in physiological and pathological contexts, ultimately advancing its application in precision therapy for NSCLC.

CRediT authorship contribution statement

Yikun Wang: Conceptualization, Formal analysis, Investigation, Methodology. Susu Guo: Investigation. Wanxin Guo: Formal analysis, Investigation, Methodology. Xiaoting Tian: Investigation. Yayou Miao: Investigation. Shiyu Qiu: Investigation. Xiangfei Xue: Methodology. Yongjie Wang: Methodology. Jiangtao Cui: Methodology. Xin Xu: Methodology. Jiayi Wang: Supervision, Writing – review & editing. Xiao Zhang: Conceptualization, Funding acquisition, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.