A non-canonical vitamin K cycle is a potent ferroptosis suppressor

Ferroptosis, a non-apoptotic form of cell death marked by iron-dependent lipid peroxidation1, has a key role in organ injury, degenerative disease and vulnerability of therapy-resistant cancers2. Although substantial progress has been made in understanding the molecular processes relevant to ferroptosis, additional cell-extrinsic and cell-intrinsic processes that determine cell sensitivity toward ferroptosis remain unknown. Here we show that the fully reduced forms of vitamin K—a group of naphthoquinones that includes menaquinone and phylloquinone3—confer a strong anti-ferroptotic function, in addition to the conventional function linked to blood clotting by acting as a cofactor for γ-glutamyl carboxylase. Ferroptosis suppressor protein 1 (FSP1), a NAD(P)H-ubiquinone reductase and the second mainstay of ferroptosis control after glutathione peroxidase-44,5, was found to efficiently reduce vitamin K to its hydroquinone, a potent radical-trapping antioxidant and inhibitor of (phospho)lipid peroxidation. The FSP1-mediated reduction of vitamin K was also responsible for the antidotal effect of vitamin K against warfarin poisoning. It follows that FSP1 is the enzyme mediating warfarin-resistant vitamin K reduction in the canonical vitamin K cycle6. The FSP1-dependent non-canonical vitamin K cycle can act to protect cells against detrimental lipid peroxidation and ferroptosis.

Unrestrained iron-dependent lipid peroxidation is the common downstream cellular event leading to rupture of cellular membranes and ferroptosis 7 . Cells have evolved a number of highly efficient redox systems that counteract uncontrolled lipid peroxidation, such as selenium-dependent glutathione peroxidase-4 (GPX4), the FSP1-ubiquinone pathway, and the biopterin-dihydrofolate reductase system 4,[8][9][10][11][12] . In addition, cells and tissues harness vitamin E (comprising both tocopherols and tocotrienols), Nature's premier lipophilic radical-trapping antioxidants 13 (RTA), to protect them from overwhelming lipid peroxidation and ferroptosis 7 . Vitamin E has also been shown to rescue certain tissues, including liver, endothelium, CD8 + T cells and hematopoietic stem cells, from the deleterious consequences induced by the tissue-specific disruption of the key ferroptosis regulator GPX4 [14][15][16] .

Vitamin K is a potent anti-ferroptotic compound
To interrogate whether there are other systems besides the aforementioned intrinsic and extrinsic mechanisms that efficiently prevent ferroptosis, we systematically studied a number of naturally available vitamin compounds in mouse embryonic fibroblasts with tamoxifen (TAM)-inducible deletion of Gpx4 (referred to as Pfa1 cells 8 ) (Extended Data Fig. 1a). Notably, besides α-tocopherol (α-TOH), the most biologically active form of vitamin E, only the three forms of vitamin K, phylloquinone, menaquinone-4 (MK4) and menadione, rescued cells from ferroptosis induced by Gpx4 deletion (Fig. 1a,b). Phylloquinone is obtained mostly from leafy green vegetables, and can be converted to MK4 in the body, whereas menadione is a synthetic variant. The anti-ferroptotic activity of vitamin K was not only limited to mouse fibroblasts, as it also prevented ferroptosis in the human cancer cell lines A375 and 786-O that lack GPX4 expression ( Fig. 1c and Extended Data Fig. 1b). Phylloquinone, MK4 and menadione also efficiently rescued cells from ferroptosis triggered by well-established ferroptosis inducers 17 including a GPX4 inhibitor (1S,3R)-RSL3 (RSL3) in fibrosarcoma HT-1080 cells ( Fig. 1d and Supplementary Videos 1 and 2), as well as in other cancer and non-cancer cell lines (Extended Data Fig. 1c,d).
In addition, all three vitamin K forms prevented glutamate-induced neuronal ferroptosis, whereas they failed to protect against other types of cell death, except protection of pyroptosis by menadione 18 (Extended Data Fig. 1e,f). Phylloquinone and MK4 showed no cellular toxicity up to 100 µM, although high doses of menadione (over 10 µM) showed signs of toxicity, probably owing to the generation of reactive oxygen species, as reported 19 (Extended Data Fig. 1g). Menadione, which lacks an aliphatic sidechain, also rescued cells from ferroptosis-albeit with lower efficacy-whereas dimethylmenadione, a redox-inactive form, did not prevent ferroptosis (Extended Data Fig. 1h). This suggests that the quinone head group is a structural requirement for the anti-ferroptotic function.
As iron-dependent lipid peroxidation is the hallmark of ferroptosis, we evaluated the levels of lipid peroxidation by performing high-resolution liquid chromatography-mass spectrometry (LC-MS)based epilipidomics analysis, staining cells with BODIPY 581/591 C11 and Liperfluo, and determining malondialdehyde concentrations   Fig. 2a-c). These studies showed that all three forms of vitamin K efficiently suppressed lipid peroxidation, and through a mechanism independent of any iron-chelating effect (Extended Data Fig. 2d), clearly prevented the RSL3-induced formation of oxidized lipid species, of which the most prominent were long-chain oxidized phosphatidylethanolamine and phosphatidylethanolamine plasmalogens ( Fig. 1e and Extended Data Fig. 2a). These data demonstrate that the vitamin K family of compounds acts as potent anti-ferroptotic agents.

Vitamin K protects tissues from ferroptosis
We then tested whether vitamin K can also prevent ferroptosis in vivo using mice with genetic deletion of Gpx4 and pathological models, in which ferroptosis results in tissue injury. We focused on MK4 as it was the most efficacious derivative (Fig. 1a,c and Extended Data Fig. 1b). First, we treated mice with TAM-inducible deletion of Gpx4 in hepatocytes (referred to as Alb-creER T2 ;Gpx4 fl/fl mice) (Fig. 1g). As reported previously, hepatocyte-specific Gpx4-knockout (KO) mice died soon after treatment as a result of widespread liver necrosis when the standard diet was switched to a vitamin E-deficient diet 15 (Extended Data Fig. 3a). Of note, treatment with a supra-nutritional level of MK4 extended the survival time of Alb-creER T2 ;Gpx4 fl/fl mice under vitamin E-deficient conditions and robustly protected against related pathologic changes and lipid peroxidation in the liver (Fig. 1h-j and Extended Data Fig. 3b-d).
To address whether MK4 might also be protective in a model of ischaemia-reperfusion injury 10,20 , we treated C57BL/6 mice with MK4 before liver or kidney ischaemia-reperfusion injury. Pre-treatment with MK4 in the mouse liver ischaemia-reperfusion injury model ameliorated liver injury with reduced lipid peroxidation, decreased hepatocyte cell death and diminished infiltration of inflammatory cells (Extended Data Fig. 4a). In the kidney ischaemia-reperfusion injury model, pre-treatment of MK4 also conferred protection against tissue damage reflected by a reduced number of terminal deoxynucleotidyl transferase dUTP nick end labelling (TUNEL)-positive tubular cells, reduced expression of the tubular damage marker kidney injury molecule-1 (KIM-1) and improved kidney function (Extended Data Fig. 4b). We thus conclude that a pharmacological dose of vitamin K has a potent anti-ferroptotic effect preventing against cell death in relevant in vivo models of ferroptosis.

FSP1 maintains VKH 2 to act as an RTA
Vitamin K is a redox-active naphthoquinone, which is converted to its corresponding hydroquinone (VKH 2 ) in the well-established vitamin K cycle 3 (Extended Data Fig. 4c,d). VKH 2 is reported to be a potent RTA 21 preventing lipid peroxidation 22 in addition to its canonical function as a cofactor for γ-glutamyl carboxylase (GGCX), which catalyses the carboxylation of vitamin K-dependent proteins, including coagulation factors. During the GGCX-mediated reaction, VKH 2 is oxidized to vitamin K epoxide, and then converted to vitamin K quinone by vitamin K epoxide reductase (VKOR), whose activity is inhibited by warfarin. The reduction of vitamin K to VKH 2 is also mediated by VKOR or an alternative warfarin-resistant pathway, catalysed by NAD(P)H-dependent vitamin K reductase activity 3,23 (Extended Data Fig. 4d). The identity of the warfarin-resistant vitamin K reductase remains unknown despite it was first described more than half a century ago 6,24 . FSP1 (encoded by the AIFM2 gene), a NAD(P)H-dependent ubiquinone oxidoreductase, is a major means of ferroptosis control, acting independently of GPX4 by catalysing the reduction of ubiquinone to its hydroquinone form, ubiquinol, consuming NAD(P)H 4,5 . Since vitamin K and ubiquinone share structural properties (Extended Data Fig. 4c), we tested whether FSP1 could act as a vitamin K reductase, producing VKH 2 to inhibit lipid peroxidation. We interrogated this possibility using in vitro assays with recombinant human FSP1 (rhFSP1). When any of the three vitamin K forms were co-incubated with rhFSP1 and NADH, NADH was consumed (Fig. 2a). In addition, when MK4 was incubated with rhFSP1 and NADH, MK4-hydroquinone was generated ( Fig. 2b and Extended Data Fig. 5a  of vitamin K. Given the instability of VKH 2 to autoxidation (Extended Data Fig. 5b,c), which complicates quantification and kinetic characterization of the enzyme, we synthesized a pro-fluorescent vitamin K analogue to enable direct monitoring of FSP1 activity using fluorescence ( Fig. 2c and Extended Data Fig. 6a-d). This vitamin K analogue proved to be a good substrate for FSP1 with similar Michaelis constant (K M ) and maximum reaction velocity (v max ) than a pro-fluorescent ubiquinone analogue. Of note, FSP1-mediated enzymatic activity could be prevented by the FSP1 inhibitor iFSP1 4 , but was insensitive to warfarin (Extended Data Fig. 6e).
We evaluated lipid radical-trapping activity by FSP1-mediated vitamin K reduction using the fluorescence-enabled inhibited autoxidation (FENIX) assay 25 , in which liposomal lipid peroxidation is monitored fluorometrically by the competitive oxidation of STY-BODIPY (Extended Data Fig. 7a). Although phylloquinone, MK4 and menadione themselves are not inhibitors of lipid peroxidation, in the presence of rhFSP1 and NADH, the three vitamin K forms efficiently suppressed lipid peroxidation (Fig. 2d). NADH itself did not suppress lipid peroxidation (Extended Data Fig. 7b), but the supply of NADH clearly extended the duration of the inhibited period of each of the three forms of vitamin K with FSP1, indicating that NADH functioned as the stoichiometric reductant (Fig. 2d). The oxidation rate depended on the concentrations of NADH, vitamin K and FSP1 (Fig. 2d-f and Extended Data Fig. 7c,d), supporting the notion that VKH 2 produced by the FSP1-catalysed reduction of vitamin K is the active RTA preventing lipid peroxidation. MK4 and phylloquinone, which both possess side chains, showed higher initial rates of oxidation, but inhibited the oxidation far longer in the presence of FSP1 and NADH (inhibition rate constant (k i ) = 5.4 × 10 3 and k i = 1.5 × 10 3 M −1 s −1 , respectively); by contrast, menadione, the least lipophilic form, displayed the fastest radical-trapping kinetics, as indicated by the most suppressed initial rate combined with the shortest inhibition period (k i = 1.1 × 10 4 M −1 s −1 Extended Data Fig. 7i). Thus, poorer dynamics of phylloquinone and MK4 in the lipid bilayer owing to the lengthy side chains may give rise to their localization within the lipid membrane, suppressing their autoxidation and the consumption of reducing equivalents from NADH or NADPH (Extended Data Fig. 7e). Notably, in the presence of FSP1, the RTA activity of phylloquinone was similar to that of ubiquinone, whereas MK4 was much more efficient (compare Fig. 2d-f and Extended Data Fig. 7f). Unlike menadione, dimethylmenadione did not show any RTA activity (Extended Data Fig. 7g). It follows that the naphthoquinone head group confers RTA function to vitamin K, which is sufficient to prevent lipid peroxidation and subsequent ferroptosis.
Given the fact that ubiquinol can work in concert with α-TOH to suppress lipid peroxidation, and that we observed such a synergy enabled by FSP1 in our previous work 4 , we also investigated the combination of vitamin K derivatives with α-TOH in the presence of FSP1. In each case, the inhibition period was extended, but the oxidation rate was the same as in the presence of α-TOH alone ( Fig. 2g and Extended Data Fig. 7h,i), implying that it is the reactive RTA that is regenerated from the α-tocopheroxyl radical (α-TO•) by VKH 2 as is the case for ubiquinone and ubiquinol (Extended Data Fig. 8a). We additionally confirmed the RTA activity of VKH 2 reduced chemically or by FSP1-mediated reaction using LipiRADICALGreen (previously called NBD-Pen 26,27 ), a fluorescence probe for lipid-derived radicals (Extended Data Fig. 8b-f).

Vitamin K blocks ferroptosis via FSP1
Consistent with the observation that FSP1-mediated vitamin K reduction is responsible for RTA activity, phylloquinone and MK4 showed a diminished anti-ferroptotic effect against RSL3 in FSP1-KO cells similar to α-TOH, which can also be regenerated by FSP1 4 (Fig. 3a and Extended Data Fig. 9a-c). Reconstitution of FSP1 expression recovered the anti-ferroptotic function of phylloquinone and MK4, whereas expression of the myristoylation-defective G2A mutant of FSP1 4 did not rescue the protective effects of phylloquinone and MK4 in FSP1-KO cells ( Fig. 3a and Extended Data Fig. 9d). In line with these findings, GPX4 and FSP1 double-KO cells required higher concentrations of phylloquinone and MK4 to prevent ferroptosis than GPX4 single-KO cells (Fig. 3b). Pharmacological inhibition of FSP1 by iFSP1 4 also diminished the protective effects of phylloquinone and MK4 against ferroptosis induced by RSL3 and by genetic deletion of GPX4 (Fig. 3c,d). Treatment with warfarin and dicoumarol (an NAD(P)H quinone dehydrogenase 1 (NQO1) inhibitor) did not significantly influence the anti-ferroptotic activity of phylloquinone and MK4 ( Fig. 3c and Extended Data Fig. 9e), although warfarin suppressed the protective effect of MK4 epoxide by inhibiting the conversion to MK4 (Extended Data Fig. 9f). These findings indicate that the reduction of vitamin K by FSP1 is responsible for the anti-ferroptotic action of phylloquinone and MK4. However, even in FSP1-KO cells, high doses of phylloquinone and MK4 still prevented ferroptosis (Fig. 3c), suggesting that other mechanisms (although less efficient) may contribute to the reduction of vitamin K. Indeed, menadione can be reduced non-enzymatically by glutathione (Extended Data Fig. 5c), and enzymatically by NQO1 and/or thioredoxin reductase in addition to FSP1 28,29 . Thus, genetic deletion and pharmacological inhibition of FSP1 did not significantly influence the anti-ferroptotic effect of menadione, similar to other FSP1-independent ferroptosis inhibitors (Extended Data Fig. 9c,g).

Antidotal FSP1 averts warfarin poisoning
Warfarin is one of the most widely prescribed anticoagulant drugs worldwide. High-dose vitamin K is an effective antidote for warfarin poisoning 30,31 because sufficient input of vitamin K can provide VKH 2 for GGCX through the alternative vitamin K reduction pathway, bypassing dysfunctional VKOR by the unidentified antidotal enzyme 32 (Extended Data Fig. 4d). Since FSP1 was capable of reducing vitamin K, we tested whether FSP1 is the enzyme responsible for the warfarin-resistant alternative vitamin K reduction pathway in the canonical GGCX-VKORmediated cycle 3 . When human liver HepG2 cells were treated with MK4, it was immediately metabolized to the MK4 epoxide via the hydroquinone and then converted to MK4 again 33 . However, the conversion efficiency to the MK4 epoxide was significantly lower in the FSP1-deficient cells, especially when cells were treated with warfarin ( Fig. 4a), indicating that FSP1 is responsible for vitamin K reduction, in addition to VKOR, in the canonical cycle. We next examined the function of FSP1 in this pathway using Fsp1 +/and Fsp1 -/mice subjected to warfarin overdose in the presence or absence of MK4 treatment (Extended Data Fig. 10a-c). When a high dose of warfarin was administered, the Fsp1 -/mice treated with MK4 showed much less conversion of MK4 to MK4 epoxide ( Fig. 4b and Extended Data Fig. 10d,e) and still showed extremely prolonged prothrombin times (a parameter of vitamin K-dependent coagulation factors) in contrast to Fsp1 +/mice (Fig. 4c). Whereas almost all warfarin-treated groups had to be euthanized, mainly owing to cerebral bleeding, one Fsp1 allele was sufficient to enable complete rescue by high-dose vitamin K treatment ( Fig. 4d and Extended Data Fig. 10f,g), corroborating that FSP1 is the warfarin-resistant vitamin K reductase in the canonical vitamin K cycle.

Discussion
Long before the term ferroptosis was introduced in 2012 1 , an antioxidative effect of vitamin K was reported 22,34 , although its mechanism remained obscure. Here, we show that vitamin K confers robust anti-ferroptotic activity via its reduced form, VKH 2 . We further demonstrate that the previously recognized ferroptosis suppressor FSP1 is the vitamin K reductase that sustains a warfarin-insensitive non-canonical vitamin K cycle that suppresses ferroptosis by maintaining VKH 2 at the expense of NAD(P)H to prevent lipid peroxidation (Fig. 4e). Furthermore, our data unveil FSP1 as the antidotal enzyme overcoming warfarin poisoning. Phylloquinone and menaquinone are electron carriers used in plants and bacteria, respectively, whereas eukaryotes use ubiquinone. FSP1 thus reduces both of the electron transfer quinones, generating RTAs. Considering the evolution of life, when environmental oxygen concentrations increased after the great oxidation event in primordial Earth, it appears that menaquinone was substituted by ubiquinone as an electron carrier owing to its higher redox potential and increased abundance compared with vitamin K 35 . Since ferroptosis is an evolutionarily conserved cell death mechanism in diverse species  . One-way ANOVA with Dunnett's test (a), two-tailed t-test using log-transformed values (b), two-tailed t-test (c) and log-rank test (d). e, Right, graphical abstract depicting the anti-ferroptotic function of vitamin K via FSP1-mediated reduction and lipid radical-trapping activity, thus constituting a non-canonical vitamin K redox cycle. Left, FSP1 also functions as the warfarin-resistant vitamin K reduction pathway overcoming warfarin poisoning in the canonical cycle. VK, vitamin K; VKH 2 , vitamin K hydroquinone; Glu, glutamate; Gla, γ-carboxyglutamate; PLOO•, phospholipid peroxyl radical; PLO•, phospholipid alkoxyl radical. The illustration of the vessel was created using BioRender.com. ranging from prokaryotes and plants to mammals 36 , our findings suggest that vitamin K might be the most ancient type of naturally occurring anti-ferroptotic quinones.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-05022-3.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Cell viability assays
Cells were seeded on 96-well plates and allowed to adhere overnight. On the next day, cells were treated with cell death inducers. Vitamin K compounds were added to the medium 1 h prior to the treatment with the ferroptosis inducing agents. Warfarin, iFSP1 and dicoumarol were added during cell seeding on 96-well plates. Cell viability was assessed 24 h (RSL3, erastin, FINO 2 , ML210, ML162, glutamate and staurosporine), 48 h (17AAG and FIN56) and 60 h (BSO) after the treatment using AquaBluer (MultiTarget Pharmaceuticals) as an indicator of viable cells. The cell viability was expressed as relative values compared to the control sample, which was defined as 100%. Pfa1 cells were seeded on 96-well plates (500 cells per well) and treated in a dilution series of the compounds and 1 µM 4-OH-TAM to induce the KO of Gpx4. Cell viability of Pfa1 cells was assessed 72 h after TAM treatment. To directly monitor cell death, LDH release was used, whereby LDH activity in medium was measured using LDH Cytotoxity Detection kit (Takara Bio). human Fas ligand (30 ng ml −1 , ALX-522-020, Enzo Life Sciences) for 24 h, and Pfa1 cells (1,500 cells per well) were co-incubated with mouse TNF (10 ng ml −1 ) and BV-6 (400 nM) for 24 h. For necroptosis induction, L929 cells (10,000 cells per well) were co-incubated with mouse TNF (10 ng ml −1 ), BV-6 (400 nM) and zVAD-FMK (30 µM) for 24 h. For pyroptosis induction, LPS (1 µg ml −1 , 4h)-stimulated THP-1 cells (20,000 cells per well) were pretreated with vitamin K or MCC950 for 1 h and then incubated with nigericin (10 µM) for 2 h.
Preparation of lentiviral particles Lentiviral packaging system consisting of a transfer plasmid, psPAX2 (12260, Addgene), and pMD2.G (12259, Addgene) was co-lipofected into HEK293T cells using the X-tremeGENE HP agent (Roche). Cell culture supernatants containing viral particles were collected 48 h after the transfection and used to transduce the cell line of interest after filtration using a 0.45 µm low protein binding syringe filter.
Transient expression of the CRISPR-Cas9 system A375 and B16F10 cells were transiently co-transfected with the desired sgRNA expressing lentiCRISPR v2-blast and lentiCRISPR v2-puro using the X-tremeGENE HP agent (Roche). One day after transfection, cells were selected by incubation with puromycin (1 µg ml −1 ) and blasticidin (10 µg ml −1 ). After selection, single-cell clones were picked and knockout clones were identified by sequencing out-of-frame mutations and immunoblotting.

Overexpression and Dox-inducible expression of FSP1
Codon-optimized human FSP1 gene with a C-terminal HA tag was cloned in the expression vector p442-Neo and Dox-inducible lentivirus vector pSLIK-Neo (25735, Addgene). FSP1-KO cells were infected with VSV-G pseudotyped lentiviral particles containing the hFSP1-cloned transfer plasmids. One day after infection, cells were selected with geneticin (1 mg ml −1 ). Reconstitution of FSP1 expression was verified by immunoblotting. Dox-inducible FSP1 expression was additionally verified by immunoblotting after treatment with increased concentrations of Dox for 24 h. For determining cell viability, cells were treated with increasing concentrations of Dox overnight and maintained in medium containing the same concentration of Dox during the assay period.
Live-cell imaging HT-1080 cells (80,000 cells) were seeded on µ-Dish 35 mm low (80136, i-bidi) and incubated overnight. On the next day, the cells were treated with or without 3 µM MK4 1 h before the addition of 0.5 µM RSL3. Live-cell imaging was performed using 3D Cell Explorer and Eve software v1.8.2 (Nanolive). Images were obtained at 1 min intervals. During imaging, the cells were maintained at 37 °C and 5% CO 2 by using a temperature-controlled incubation chamber.

Liperfluo staining
Cellular lipid hydroperoxides were detected using the fluorescent probe Liperfluo (Dojindo). H9C2 cells were plated onto black, clear-bottom µClear 96-well culture plates (Greiner). After removal of the medium, the cells were incubated in HBSS containing 2 µM of Liperfluo. Subsequently, the cells were incubated with 100 µM BSO and the indicated compounds for 40 h. After the incubation, the cells were washed with HBSS and observed using a BZ-X800 fluorescence microscope (Keyence). The signal intensity per cell was measured with ImageJ software v1.53 (NIH).

Iron-chelating activity assay
Iron-chelating activity was measured by Metalloassay kit Fe (FE02M, Metallogenics). After the addition of the compounds (final concentration, 100 µM) into an iron standard solution of 200 µg dl −1 of iron(iii) nitrate, free iron levels in the solution were measured according to the protocol.
Lipids from mouse livers (approximately 150 mg wet tissue weight) were extracted according to Folch extraction method 38 . SPLASH LIPIDO-MIX (Avanti Polar Lipids, 30 µl) was added. Samples were homogenized in methanol (1 ml) by cryomilling and transferred in 10 ml glass tubes. Lysis beads were washed with methanol (400 µl) and chloroform (1,000 µl). Additional 1.8 ml chloroform was added, samples were vortexed (2 min, 2500 rpm) and incubated for 1 h at 4 °C with rotation (32 rpm). Phase separation was induced by adding water (840 µl). Samples were mixed by vortexing and incubated for 10 min 4 °C with rotation, before centrifugation (10 min, 1,000g, 4 °C). The lower, organic phase containing lipids was collected into new glass vials. For re-extraction, chloroform (2.8 ml) was added, samples vortexed, incubated (1 h, 4 °C, and 32 rpm), and centrifuged (10 min, 1,000g, 4 °C). The organic phases, combined from both extractions, were dried in a vacuum. Lipids were reconstituted in 300 µl IPA, and transferred in glass vials for LC-MS analysis. To avoid oxidation, all solvents used for lipid extraction were spiked with 0.1% (w/v) BHT and cooled on ice before use.

Generation of monoclonal antibodies against mouse Fsp1
Female Wistar rats (RjHan:Wi, age 160 days) were immunized subcutaneously and intraperitoneally with a mixture of 70 µg recombinant C-terminal-His tagged full-length mouse Fsp1 protein in 200 µl PBS, 5 nmol CpG2006 (TIB MOLBIOL), and 200 µl Incomplete Freund's adjuvant (Sigma-Aldrich). After 8 weeks, a boost without Freund's adjuvant was given intraperitoneally and subcutaneously 3 days before fusion. Fusion of the myeloma cell line P3X63-Ag8.653 (CRL-1580, ATCC) with the rat immune spleen cells was performed using polyethylene glycol 1500. After fusion, the cells were plated in 96-well plates using RPMI 1640 medium with 20% FBS, glutamine, pyruvate, non-essential amino acids and HAT media supplement (Hybri-Max, Sigma-Aldrich). Hybridoma supernatants were screened 10 days later in a flow cytometry assay for binding to c-His tagged Fsp1 protein captured via biotinylated mouse anti-His antibody (clone HIS 3D5, prepared in-house) to streptavidin beads (PolyAN). Hybridoma supernatant was incubated for 90 min with beads and Atto-488-coupled subclass-specific monoclonal mouse-anti-rat IgG. Antibody binding was analysed using ForeCyt software 8 (Sartorius). Positive supernatants were further validated by Western blotting. Selected hybridoma cells were subcloned by limiting dilution to obtain stable monoclonal cell lines.

Production of purified recombinant human FSP1
Recombinant human FSP1 protein (rhFSP1) was produced in Escherichia coli, and purified by affinity chromatography with a Ni-NTA system as described previously 4 .

FENIX assays
Liposomes were prepared from egg phosphatidylcholine (egg PC, Sigma-Aldrich) in pH 7.4 TBS buffer (25 mM, extruded to 100 nm, Chelex-100 treated) according to our previous report 25,43 . Liposomes (from the above suspension), STY-BODIPY (from a 1.74 mM stock in DMSO) and the test quinone (from appropriate stock solutions in CH 3  The plate was incubated at 37 °C for 1 min followed by another 1 min wherein it was mixed and the fluorescence (λ ex /λ em = 488/518 nm) recorded every 2 min for the duration of the experiment. For determinations of inhibition rate constants, the rate of initiation (R i ) under the exact experimental conditions was first determined from the inhibition period observed upon inclusion of PMC, for which n = 2, as a representative data trace is shown in Extended Data Fig. 7a. The R i was calculated from the expression below to yield R i = 7.81 × 10 −10 s −1 from t inh = 10,240 s, where t inh is the inhibited period. This R i was used along with the expression in Extended Data Fig. 7a to calculate the rate constants shown in Extended Data Fig. 7i .

Synthesis of 1,4-dimethoxy-2-methylnaphthalene
To a solution of 2-bromo-1,4-dimethoxy-3-methylnaphthalene 44 (dimethylmenadione) (600 mg, 2.13 mmol) in THF (20 ml) was added n-butyllithium (1.02 ml, 2.5 M in n-hexane) dropwise at −78 °C; the mixture was stirred at −78 °C for 10 min, followed by the addition of 0.5 ml water. The cooling bath was removed and reaction mixture was warmed to room temperature. Solvent was removed, and the mixture was purified by column with ethyl acetate/hexanes as the eluent. Monitoring FSP1 activity with quinone-coumarin conjugates FAD, NADH and rhFSP1 in pH 7.4 TBS buffer were added in succession to varying concentrations of vitamin K-coumarin or CoQ-coumarin conjugate in pH 7.4 TBS buffer at 37 °C (final concentration: 6 nM rhFSP1, 50 nM FAD, 200 µM NADH) and the initial rates of the reaction were obtained by monitoring the increase in the fluorescence upon the reduction of the quinone to the hydroquinone on a plate reader (λ ex /λ em = 415/470 nm). The raw fluorescence data were converted to hydroquinone concentrations using response factors of 4.64 × 10 9 RFU µM −1 (for CoQ-coumarin) and 3.40 × 10 9 RFU µM −1 (for vitamin K-coumarin) which were determined from a standard curve obtained from the maximum fluorescence recorded for various concentrations of the quinones in the presence of massive excess of either rhFSP1, FAD or NADH.

LipiRADICAL Green assay
LipiRADICAL Green assay was performed according to a previous report 27  FSP1 activity assay by measuring NADH consumption FSP1 enzymatic assay was performed as described with a minor modification 4 . NADH consumption was measured at 340 nm using 100 µl of enzyme reactions in PBS pH 7.4 on a 96-well plate. Enzyme reactions contained 150 nM rhFSP1, 200 µM NADH (freshly prepared in water) and 300 µM of different substrate candidates (phylloquinone, MK4, and menadione). A Spectra Max M5 Microplate Reader (Molecular devices) was used to determine the absorbance at 340 nm every 30 s. Reactions without NADH or without enzyme were used to normalize the results.

Chemical reduction of menadione
Menadione and menadiol (300 µM) were incubated with 1 mM DTT or 10 mM GSH in water at room temperature for 5 min, and then measured by absorbance spectrum ranging from 200 to 450 nm by using a Spectra Max M2e (Molecular Devices). Background control (in blank well) of absorbance values was subtracted from each individual absorbance value.

Quantification of cellular MK4 and MK4 epoxide levels
HepG2 cells (1 × 10 6 cells per well) were seeded on 6 well plates. On the next day, medium was replaced with fresh medium with or without warfarin (10 µM). On the following day, cells were incubated in the presence or absence of MK4 (3 µM) for 7 h. After washing with PBS three times, cells were trypsinized and collected. Cell pellets were suspended in 400 µl PBS, supplemented with 20 µl of MK4-d 7 (2 ng µl −1 in ethanol, 25709, Cayman) as internal standard, and sonicated for 30 s with a sonication probe (Bronson Sonifer). In this procedure, 10 µl of cell lysate was analysed for protein determination with a BCA protein assay (Pierce BCA Protein Assay Kit, Thermo Fisher). Extraction of vitamin K and its metabolites from cells was performed as reported 33 . Four-hundred microlitres ethanol and 1.2 ml hexane were added to the cell lysate (in PBS, 400 µl) followed by shaking for 5 min. Samples were centrifuged at 1,000g for 5 min, and the upper organic layer was collected. Re-extraction of the remaining aqueous phase was performed by addition of 150 µl ethanol and 450 µl hexane with subsequent vortexing. Samples were centrifuged at 1,000g for 5 min. Collected organic layers were combined, spiked with 20 µl of phylloquinone (2 ng µl −1 in ethanol) as recovery standard and evaporated under reduced pressure. Dried extracts were resuspended in 30 µl ethanol. Quantification of the target analytes (MK4 and MK4 epoxide) was achieved using an Agilent 5890 Series II gas chromatograph (GC) coupled with a Thermo Finnigan SSQ7000 single quadrupole mass spectrometer (MS). Chromatographic separation was carried out on a Restek Rtx-5Sil MS column (30 m × 0.25 mm internal diameter × 0.25 µm film thickness). Two microlitres of each sample was injected in splitless mode using helium as carrier gas at a constant pressure of 16 psi. The injection temperature was 280 °C. Initial column temperature was 90 °C held for 1.5 min, increased to 220 °C at a rate of 20 °C min −1 , followed by a second ramp to 320 °C at a rate of 10 °C min −1 and held for 10 min. The mass spectrometer was operated in negative chemical ionization mode and the masses of the negative molecular ions were registered in single ion monitoring mode.

Quantification of MK4 and MK4 epoxide in mouse samples
Blood samples of mice were collected by bleeding from the retroorbital plexus into citrate-treated tubes. After centrifugation (3,000g for 10 min), plasma samples were obtained and stored −80 °C until analysis. Liver tissues were collected from mice after transcardiac perfusion with 10 ml PBS, snap-frozen into liquid nitrogen and stored at −80 °C. For sample preparation of plasma, 100 µl aliquots of plasma were transferred into glass tubes, spiked with 20 ng of MK4-d 7 and briefly mixed. Next, 2 ml ethanol, 4 ml hexane and 100 µl water containing butylated hydroxytoluene (0.1 %, w/v) were added. After vigorously mixing for 5 min, the samples were centrifuged at 2,200g for 5 min at 4 °C. The upper layer was transferred into a clean glass tube, and the samples were then re-extracted by the addition of an equal volume of hexane. Both Article supernatants were collected and evaporated in vacuo. The samples were dissolved in 2 ml hexane and loaded onto silica columns. For sample preparation of tissues, the tissues (kidney and liver) were weighed, transferred to lysing matrix tubes containing stainless steel beads (MP Biomedicals), and then thoroughly homogenized in 1 ml ethanol containing 20 ng of MK4-d 7 . The tissue homogenates were transferred into glass tubes using glass Pasteur pipettes. Following the addition of 6 ml acetone containing BHT (0.1 %, w/v), the homogenates were thoroughly mixed using a Ohaus Multi-Tube Vortex mixer, for 5 min at 2,500 rpm, allowed to stand for 5 min, and centrifuged at 2,200g for 5 min at 4 °C. This procedure was repeated three times. Supernatants were collected and evaporated in vacuo. The samples were dissolved in 2 ml water and 6 ml hexane containing BHT (0.

Hepatocyte-specific inducible Gpx4-KO mice
To generate mice with a TAM-inducible hepatocyte-specific deletion of Gpx4 (Alb-creER T2 ;Gpx4 fl/fl ), Gpx4 fl/fl mice were first crossbred with Alb-creER T2 mice 48 (kindly provided by P. Chambon) to yield Alb-creER T2 ; Gpx4 fl/+ mice. These were then crossed with Gpx4 fl/fl mice to generate Alb-creER T2 ;Gpx4 fl/fl mice and respective controls. To achieve inducible disruption of the floxed Gpx4 alleles, mice were intraperitoneally injected with 2 mg TAM (dissolved in Miglyol 812, Caelo) on two consecutive days. Animals were equally distributed between sex and weight and were typically 8-10 weeks of age. For pharmacological treatment, vehicle or MK4 (100 mg kg day −1 dissolved in Miglyol, twice daily) was intraperitoneally administrated to the mice each day starting from 2 days before the first TAM injection until the completion of the study. The diet was changed from a standard diet (containing 143 mg kg −1 vitamin E, no. 1314 Fortified, Altromin) to a vitamin E-deficient diet (containing <7 mg kg −1 vitamin E, E15314-247, ssniff Spezialdiäten) at the timing of the first TAM injection. When animals reached the humane end point, they were immediately euthanized. For the end point analysis, the mice were euthanized 7 days after the first TAM injection, and the plasma and tissues were collected. Serum ALT were measured by AU480 chemistry analyser (Beckman Coulter). For the pharmacokinetic study of MK4, samples of plasma, liver and kidney were collected from Gpx4 fl/fl mice 0, 1, 3, 6 and 24 h after an intraperitoneal injection of MK4 (200 mg kg −1 dissolved in Miglyol).

Liver ischaemia-reperfusion injury model in mice
Eight to 10-week-old male C57BL/6J mice, provided by Charles River (Germany), were fed a standard diet (containing 135 mg kg −1 vitamin E, ssniff Spezialdiäten) and underwent liver ischaemia-reperfusion injury as described previously 10 . In brief, mice were aneasthetized with xylazine/ketamine and shaved at their front. After opening the abdominal cavity an atraumatic clip was placed across the portal vein, hepatic artery and bile duct, just above branching to the right lateral lobe. After 90 min of ischaemia, the clamp was removed and the liver was reperfused. Mice were euthanized 24 h following transient ischaemiareperfusion and blood and tissues were collected. MK4 (200 mg kg −1 dissolved in Miglyol 812) or vehicle was injected intraperitoneally 24 h and 1 h before the onset of ischaemia. Serum ALT was measured using a Dimension 1500 Vista Analyzer (Siemens). Calculation of the necrotic/ damaged areas (% of the whole section minus the major vessels) in the haematoxylin and eosin-stained sections were performed in a blinded manner using ZEISS Axio Vision software AxioVs v4.9 (Carl Zeiss).

Kidney ischaemia-reperfusion injury model in mice
Eight-to twelve-week-old male C57BL/6N mice (Charles River), were fed a standard diet (containing 135 mg kg −1 vitamin E, V1534-300, ssniff Spezialdiäten) and underwent renal ischaemia-reperfusion injury as described previously 49 . In brief, bilateral renal pedicle clamping was performed via a midline abdominal incision for 36 min. Throughout the surgical procedure, the body temperature was maintained between 36 and 37 °C. After removal of the clamps, the abdomen was closed allowing restoration of blood flow as also visually observed. Sham-operated mice underwent the identical surgical procedures, except clamping of renal pedicles. All mice were killed 48 h after the reperfusion. All ischaemia-reperfusion experiments were performed in a double-blinded manner. MK4 (200 mg kg −1 dissolved in corn oil) or vehicle was injected intraperitoneally 1 h before the onset of ischaemia. Serum creatinine and urea were measured in the Institute for clinical chemistry of the University Hospital Dresden (Germany). Kidney tissue damage was quantified by two researchers in a double-blind manner on a scale ranging from 0 (unaffected tissue) to 10 (severe organ damage). The following parameters were chosen as indicative of morphological damage to the kidney after ischaemia-reperfusion injury: brush border loss, red blood cell extravasation, tubule dilatation, tubule degeneration, tubule necrosis, and tubular cast formation. These parameters were evaluated on a scale of 0-10, which ranged from not present (0), mild (1-4), moderate (5 or 6), severe (7 or 8), to very severe (9 or 10). For the scoring system, tissues were stained with periodic acid-Schiff (PAS), and the degree of morphological involvement in renal failure was determined using light microscopy.  For anti-cleaved caspase-3 staining, Histofine Simple Stain MAX PO (R) Anti-Rabbit (Nichirei) was used as secondary antibody. The sections were visualized with nickel-enhanced 3,3′-diaminobenzidine (DAB, SK-4100, Vector Laboratories) for anti-GPX4 and KIM-1, or DAB and counterstaining with Mayer's Hematoxylin for anti-4HNE and cleaved caspase-3. TUNEL staining was performed using the ApopTag peroxidase in situ apoptosis detection kit (Millipore). To reduce false-positive signals, the TdT enzyme was diluted 1:16 in reaction buffer for preparation of the working solution. Gr-1 + cells were immunohistochemically stained on acetone-fixed frozen liver sections. Dried sections were blocked with 10% goat serum for 1 h, and then incubated with anti-Gr-1-FITC antibody (0.5 mg ml −1 , 553127, BD Pharmingen) for 30 min at room temperature. The sections were treated with goat anti-Rat Alexa Fluor 488 IgG (H+L) (1:500, A-11006, Invitrogen) and DAPI (5 mg ml −1 ) for visualization. Gr-1 + cells were counted per high-power field (HPF) (2,000× magnification; five HPF per slide). A blinded scientist received the slides randomly and performed all cell counting procedures.

Quantification and statistical analysis
Statistical information for individual experiments can be found in the corresponding figure legends. Values are presented as mean ± s.d. Statistical comparisons between groups were analysed for significance by two-tailed Student's t-test, one-way ANOVA with Dunnett's post hoc test. Survival analysis was done according to the log-rank test. Results were considered significant at P <0. 05. Statistical analyses were conducted using GraphPad Prism 9 (GraphPad Software) and JMP15 (SAS Institute) software.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability
Epilipidomics data are available at MASSIVE (https://massive.ucsd. edu/) under accession number MSV000089489. Gel source images are shown in Supplementary Fig. 1. Other data are available from the corresponding author upon reasonable request. Source data are provided with this paper.

Corresponding author(s): Marcus Conrad
Last updated by author(s): May 21, 2022 Reporting Summary Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.

Data
Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A description of any restrictions on data availability -For clinical datasets or third party data, please ensure that the statement adheres to our policy All data are available within the Article and the Supplementary Information, and from the corresponding author on reasonable request. Gel source images are shown in Supplementary Fig. 1. All source data are provided with this paper. Epilipidomics data are available at MASSIVE (https://massive.ucsd.edu/) under accession number MSV000089489.