Ferroptosis: A Regulated Cell Death Nexus Linking Metabolism, Redox Biology, and Disease

Cell death is critical in diverse aspects of mammalian development, homeostasis, and disease, and it is tightly integrated with other biological processes. The Nomenclature Committee on Cell Death (NCCD) recognizes a distinction between accidental cell death , which is caused by severe physical, chemical and mechanical insults and cannot be reversed by molecular perturbations, and regulated cell death , which can be modulated pharmacologically and genetically, and is therefore under the control of specific intrinsic cellular mechanisms. Moreover, the NCCD defines that programmed cell death is a subset of regulated cell death that is predestined to occur in normal physiological contexts such as development ( Galluzzi et al., 2015 ). Caspase-dependent apoptosis was the first regulated and programmed form of cell death to be characterized at the molecular level. In recent years, there has been a growing appreciation for the importance of regulated cell death mechanisms beyond apoptosis in explaining the molecular processes that control the demise of cells ( Conrad et al., 2016 ), although the extent to which these other forms of regulated cell death are also programmed to occur under normal physiological conditions remains largely unknown. We focus in this Primer on ferroptosis, an iron-dependent form of regulated cell death that involves lethal, iron-catalyzed lipid damage. As described below, ferroptotic death is a form of regulated cell death, as it is dramatically modulated by pharmacological perturbation of lipid repair systems involving glutathione and GPX4, and is dependent on a set of positive-acting enzymatic reactions, including biosynthesis of PUFA-containing phospholipids, which are the substrates of pro-ferroptotic lipid peroxidation products, by ACSL4 and LPCAT3, and selective oxygenation of PUFA-phosphatidylethanolamines by lipoxygenases. The term ferroptosis was coined in 2012 ( Dixon et al., 2012 ) to describe the form of cell death induced by the small molecule erastin, which inhibits the import of cystine, leading to glutathione depletion and inactivation of the phospholipid peroxidase glutathione peroxidase 4 (GPX4) ( Yang et al., 2014 ). GPX4 converts potentially toxic lipid hydroperoxides (L-OOH) to non-toxic lipid alcohols (L-OH) ( Ursini et al., 1982 ) ( Figure 1 ). Inactivation of GPX4 through depletion of GSH with erastin, or with the direct GPX4 inhibitor (1 S ,3 R )-RSL3 (hereafter referred to as RSL3), ultimately results in overwhelming lipid peroxidation that causes cell death ( Table 1 ). Thus, suppression of lipid peroxidation is required in mammalian cells to prevent the onset of ferroptosis, which may occur by default under conditions where GPX4 is inactivated. Since the initial description of this process, additional compounds and regulatory mechanisms have been identified ( Xie et al., 2016a ; Yang and Stockwell, 2016 ) ( Tables 1 and 2 ), and this process has been implicated in a variety of pathological contexts and therapeutic strategies, described below.

Reduced glutathione (γ-L-glutamyl-L-cysteinylglycine, GSH) is an essential intracellular antioxidant synthesized from glutamate, cysteine and glycine in two steps by the ATP-dependent cytosolic enzymes glutamate-cysteine ligase (GCL) and glutathione synthetase (GSS) ( Figure 1 ); the rate of glutathione synthesis is limited by cysteine availability. Early studies identified extracellular cysteine and cystine as essential for growth of HeLa and other cells in culture ( Eagle, 1955 ). Cysteine is required for cell growth, and to prevent cell death: human fibroblasts cultured in cystine-free medium die due to glutathione depletion, and this death is prevented by the lipophilic antioxidant α-tocopherol ( Bannai et al., 1977 ). In addition, iron chelators, such as deferoxamine, prevent this cell death ( Murphy et al., 1989 ). These results established the concept that extracellular cystine and intracellular cysteine are required to maintain the biosynthesis of glutathione and to suppress a type of cell death in mammalian cells that is also preventable by treatment with iron chelators or lipophilic antioxidants. Glutathione levels were observed to be key regulators of viability in some contexts in plants as well: abortion of premature pollen due to abnormal microspore formation in the sterile rice plant Oryza sativa L was found to involve depletion of GSH and NADPH during meiosis, and consequent accumulation of ROS and cell death, which is intriguingly reminiscent of ferroptosis ( Wan et al., 2007 ). However, when these historical studies were performed, a framework for ferroptosis was lacking, and these observations were not yet interpreted as evidence for a unique form of regulated cell death.

Lipid metabolism is also intimately involved in determining cellular sensitivity to ferroptosis. PUFAs, which contain bis-allylic hydrogen atoms that can be readily abstracted, are susceptible to lipid peroxidation and are necessary for the execution of ferroptosis ( Yang et al., 2016 ). Thus, the abundance and localization of PUFAs determine the degree of lipid peroxidation that occurs in cells, and hence the extent to which ferroptosis is operative. Free PUFAs are substrates for synthesis of lipid signaling mediators, yet they must be esterified into membrane phospholipids and undergo oxidation to become ferroptotic signals ( Kagan et al., 2017 ). Lipidomic studies suggest that phosphatidylethanolamines (PEs) containing arachidonic acid (C20:4) or its elongation product, adrenic acid (C22:4), are key phospholipids that undergo oxidation and drive cells towards ferroptotic death ( Doll et al., 2017 ; Kagan et al., 2017 ). Therefore, formation of coenzyme-A-derivatives of these PUFAs and their insertion into phospholipids are necessary for the production of ferroptotic death signals. This is another potential point of regulation of ferroptosis, and future investigations may suggest physiological contexts in which ferroptosis is triggered or blocked by modulating enzymes involved in the biosynthesis of PUFA-containing membrane phospholipids. For example, two enzymes, ACSL4 and LPCAT3 ( Table 2 ), are involved in the biosynthesis and remodeling of PUFA-PEs in cellular membranes. Loss of these gene products depletes the substrates for lipid peroxidation and increases resistance to ferroptosis ( Dixon et al., 2015 ; Doll et al., 2017 ; Kagan et al., 2017 ; Yuan et al., 2016b). Conversely, cells that are supplemented with arachidonic acid or other PUFAs are sensitized to ferroptosis ( Yang et al., 2016 ). Hydroperoxy derivatives of PUFA-PEs also cause ferroptotic death when added to cells with inactivated GPX4 ( Kagan et al., 2017 ). Intriguingly, proper development of the fetal immune system in humans depends on adequate PUFA dietary intake, suggesting a possible role for ferroptosis in this process, although other explanations are also possible ( Enke et al., 2008 ). Enzymatic effectors, such as non-heme, iron-containing proteins, including lipoxygenases (LOXs), can mediate ferroptotic peroxidation ( Kagan et al., 2017 ; Seiler et al., 2008 ; Yang et al., 2016 ). Free PUFAs, rather than PUFA-containing phospholipids, are the preferred substrates of LOXs ( Kuhn et al., 2015 ); PE phospholipids can, however, form a non-bilayer arrangement (van den Brink-van der Laan et al., 2004) that may facilitate pro-ferroptotic oxidation of PUFA-containing PE phospholipids by LOXs, rather than free PUFAs. Indeed, genetic depletion of LOXs protects against erastin-induced ferroptosis ( Yang et al., 2016 ), suggesting LOXs contribute to ferroptosis. Several ferroptosis inhibitors, including members of the vitamin E family (tocopherols and tocotrienols) and flavonoids, can inhibit LOX activity in some contexts ( Khanna et al., 2003 ; Xie et al., 2016b). Some lipoxygenases are required for normal embryonic development in vertebrates; for example, 12S-lipoxgenase is essential for the development of numerous tissues in zebrafish, suggesting that ferroptosis could be involved in these developmental processes, although further study is required to test this possibility ( Haas et al., 2011 ). The execution phase of ferroptosis may be a direct result of lipid peroxidation. Lipid peroxides decompose into reactive derivatives, including aldehydes and Michael acceptors, which can react with proteins and nucleic acids ( Gaschler and Stockwell, 2017 ). The hypothesis that these reactive intermediates are responsible for cell death is supported by the observation that a cell line selected for erastin resistance showed several-hundred-fold upregulation of AKR1C family genes ( Dixon et al., 2014 ), which encode aldoketoreductases that, among other functions, reduce reactive end-products of lipid peroxidation to unreactive compounds ( MacLeod et al., 2009 ).

Several other metabolic pathways modulate cellular sensitivity to ferroptosis. The mevalonate pathway leads to the production of coenzyme Q10 (CoQ 10 ), which is required to shuttle electrons as part of the mitochondrial electron transport chain; however, this function is not relevant to ferroptosis ( Dixon et al., 2012 ). Rather, CoQ 10 moonlights as an endogenous inhibitor of ferroptosis by serving an antioxidant function in membranes ( Shimada et al., 2016b ). The ferroptosis-inducing compound FIN56 depletes CoQ 10 by modulating squalene synthase activity (SQS), which in part drives accumulation of lethal lipid peroxidation ( Shimada et al., 2016b ). Statin drugs, which inhibit HMG CoA reductase, the rate-limiting enzyme of the mevalonate pathway, sensitize cells to ferroptosis, presumably by depleting CoQ 10 , and possibly by also inhibiting downstream tRNA isopentenylation via TRIT1, which is required for the biosynthesis of GPX4 ( Fradejas et al., 2013 ; Shimada et al., 2016b ; Viswanathan et al., 2017 ). NADPH and selenium abundance also impact ferroptosis sensitivity. NADPH is an essential intracellular reductant needed to eliminate lipid hydroperoxides. Indeed, NADPH levels are a biomarker of ferroptosis sensitivity across many cancer cell lines ( Shimada et al., 2016a ). Selenium is required for the biosynthesis of GPX4, which has an active-site selenocysteine ( Cardoso et al., 2017 ). Thus, selenium supplementation promotes ferroptosis resistance, while selenium depletion promotes ferroptosis sensitivity, presumably by modulating GPX4 abundance and activity ( Cardoso et al., 2017 ). A number of other genes have been implicated as modulators of sensitivity to ferroptosis or markers of ferroptosis in diverse contexts: SAT1 , which lies downstream of p53 ( Ou et al., 2016 ) is involved in polyamine metabolism; kiss of death (KOD) in Arabidopsis ( Distefano et al., 2017 ) and FANCD2 in bone marrow stromal cells are induced during ferroptosis ( Song et al., 2016 ); TTC35 , CS , ATP5G3 , and RPL8 are involved in diverse processes in human cancer cells and suppress erastin-induced ferroptosis upon knockdown ( Dixon et al., 2012 ). The NRF2 transcription factor promotes resistance to ferroptosis ( Sun et al., 2016b ): NRF2 controls the expression of AKR1C , the metal-binding protein MT-1G ( Sun et al., 2016a ), and other antioxidant genes, the expression of key proteins in iron signaling, including ferritin and ferroportin, and the expression of enzymes in the pentose phosphate pathway, which generate the bulk of cellular NADPH. Additionally, genes encoding proteins responsible for glutathione synthesis, including SLC7A11, GCLC/GLCM, and GSS are NRF2 target genes ( Kerins and Ooi, 2017 ).

Cancer cells are susceptible to perturbations of thiol metabolism, whereas oxidative stress via excess iron is associated with carcinogenesis ( Toyokuni et al., 2017 ). Agents that inhibit cystine uptake via the cystine/glutamate antiporter (system x c - ), such as sulfasalazine, arrest tumor growth and can induce ferroptosis in some circumstances ( Dixon et al., 2014 ; Toyokuni et al., 2017 ). Likewise, direct depletion of cystine from plasma using an engineered cystine-degrading enzyme conjugate ( i.e., cyst(e)inase) arrests tumor growth and triggers cell death ( Cramer et al., 2017 ). Agents that conjugate to glutathione, such as APR-246, as well as chemical or genetic inhibition of glutathione biosynthesis, disrupt tumor cell growth and induce a ferroptosis-like form of cell death ( Liu et al., 2017 ). Typically, elevated levels of ROS are detected in response to perturbation of cysteine or glutathione metabolism and, where tested, cell death is prevented by antioxidant treatment. Erastin and RSL3 were originally identified in phenotypic screens for compounds that are selectively lethal to engineered tumor cells. Improved analogs of erastin with increased solubility, selectivity, and potency have been created, and some have shown efficacy in xenograft tumor studies ( Yang et al., 2014 ). Nanoparticles that induce ferroptosis have also demonstrated efficacy in xenograft studies ( Kim et al., 2016 ). The tumor suppressor p53 has been reported to repress SLC7A11 , a component of system x c - , thus inducing ferroptosis in some contexts ( Jiang et al., 2015 ). In addition, CD44v, a cancer stem cell marker, associates with system x c - , and stabilizes this complex; this observation suggests that CD44v may be a biomarker for tumors that are sensitive to system x c - inhibitors that induce ferroptosis ( Toyokuni et al., 2017 ).

There are four classes of ferroptosis inducers currently in use ( Yang and Stockwell, 2016 ): (i) system x c - inhibitors (erastin and its analogs, sulfasalazine, glutamate, and sorafenib); (ii) GPX4 inhibitors, such as RSL3 and ML162; (iii) FIN56, which depletes GPX4 protein, and the lipophilic antioxidant CoQ 10 ( Shimada et al., 2016b ); and (iv) FINO2, which indirectly inhibits GPX4 activity and stimulates lipid peroxidation ( Abrams et al., 2016 ) ( Table 1 ). In addition to these canonical ferroptosis inducers, several other reagents can induce ferroptosis in some contexts: buthionine sulfoximine (BSO) depletes glutathione, and in some cases is capable of inducing ferroptosis ( Yang et al., 2014 ); CCl 4 may induce ferroptosis in the liver ( Guo et al., 2017 ), artesunate can induce ferroptosis in pancreatic cancer cells ( Eling et al., 2015 ), cisplatin induces both ferroptosis and apoptosis in several tissues ( Jennis et al., 2016 ), and a derivative of artemisinin ( Greenshields et al., 2017 ) induce both ferroptosis and apoptosis. A variety of pharmacological and genetic inhibitors of ferroptosis have also been reported ( Angeli et al., 2017 ; Conrad et al., 2016 ; Doll et al., 2017 ). Inhibitors of lipid metabolism that suppress PUFA incorporation into phospholipid membranes have been reported, such as knockdown or knockout of ACSL4 or LPCAT3, and thiazolidinediones, which inhibit ACSL4 ( Dixon et al., 2015 ; Doll et al., 2017 ). Inhibitors of lipid peroxidation, such as LOX inhibitors, also suppress ferroptosis ( Yang et al., 2016 ). Inhibitors of iron metabolism and iron chelators ( e.g., deferoxamine (DFO) and ciclopirox (CPX)) suppress ferroptosis by reducing availability of iron ( Yang and Stockwell, 2008 ). Inhibition of glutaminolysis, as noted above, suppresses ferroptosis, although the mechanism of action is not known ( Gao et al., 2015 ). The nitroxide XJB-5-131 is a potent inhibitor of ferroptosis, probably by blocking lipid peroxidation in relevant membranes ( Krainz et al., 2016 ). Similarly, ferrostatins and liproxstatins inhibit lipid peroxidation, possibly by acting as radical-trapping antioxidants ( Zilka et al., 2017 ), similarly to the lipophilic antioxidants BHT, BHA and vitamin E. In the latter case, tocotrienols are more effective than tocopherols ( Kagan et al., 2017 ). Selectively bis-allylic deuterated PUFAs suppress propagation of lipid peroxidation and have been found to suppress ferroptosis induced by RSL3 and erastin, although they are more effective against the former than the latter ( Yang et al., 2016 ). Finally, the protein synthesis inhibitor cycloheximide suppresses ferroptosis induced by system x c - inhibition ( Yagoda et al., 2007 ). Given different mechanisms of ferroptotic inhibition, the characterization of new inhibitors should be accompanied by an evaluation of antioxidant or iron-chelating activity; otherwise, the mechanism of action of inhibitors that appear to act through distinct mechanisms may be misinterpreted. For example, many LOX inhibitors and the MEK inhibitor U0126 exhibit antioxidant activity, which is probably the basis of their ability to suppress ferroptosis. In addition, the necroptosis inhibitor necrostatin-1 has off-target activity, which can suppress ferroptosis at high concentrations ( Friedmann Angeli et al., 2014 ), but this is unrelated to necroptosis. Measuring lipid peroxidation is essential for evaluating whether ferroptosis occurs in specific contexts. C11-BODIPY and Liperfluo are lipophilic ROS sensors that provide a rapid, indirect means to detect lipid ROS ( Dixon et al., 2012 ). Liquid chromatography (LC) / tandem mass spectrometry (MS) analysis has lower throughput but can be used to detect specific oxidized lipids directly ( Friedmann Angeli et al., 2014 ; Kagan et al., 2017 ). In addition, isoprostanes have been used to measure lipid peroxidation ( Milne et al., 2007 ), although not yet in the context of ferroptosis. Other useful assays for studying ferroptosis include measuring iron abundance and GPX4 activity. The former can be detected using inductively coupled plasma-MS or calcein AM quenching, as well as other specific iron probes ( Hirayama and Nagasawa, 2017 ; Spangler et al., 2016 ), while the latter can be detected using phosphatidylcholine hydroperoxide reduction in cell lysates using LC-MS ( Yang et al., 2014 ). Since ferroptosis susceptibility is so closely tied to the metabolic state of the cell, it is important in in vivo and in vitro studies to control the composition of chow and serum composition, respectively, which can vary between lots, laboratories, and experiments, by altering the levels of key ferroptotic regulators, such as selenium, iron, vitamin E, cysteine, glutathione and PUFAs in experimental samples. In addition, redox conditions vary between cell-based/organoid cultures and in vivo conditions, which is illustrated by the frequent requirement for system x c - in cell culture, whereas system

Ferroptosis is a process that has likely been discovered and re-discovered a number of times over the years before a detailed molecular understanding of this phenotype was solidified ( Dixon and Stockwell, 2013 ). Since 2012, rapid progress has been made in defining the molecular regulation of ferroptosis. We anticipate that additional regulators and biological contexts will emerge for ferroptosis in the future. We suggest here basic criteria for concluding that ferroptosis is occurring in a particular experimental system, while acknowledging that these criteria may vary depending on the mechanism by which ferroptosis is triggered. In general, ferroptosis should be suppressed by both an iron chelator (e.g., DFO or CPX) and a lipophilic antioxidant (e.g., ferrostatin, liproxstatin, vitamin E, BHT), and should involve accumulation of lipid hydroperoxides. Ferroptosis induced by system x c - inhibition should be suppressed by β-mercaptoethanol, which reduces extracellular cystine to cysteine, bypassing the requirement for system x c - for import, as cysteine is imported through other mechanisms ( Dixon, 2012 ). System x c - inhibition should result in glutathione depletion. Gene expression markers associated with cells undergoing ferroptosis include increases in CHAC1 and PTGS2 mRNA and ACSL4 protein ( Table 2 ), but these changes may not be observed in all experiments. Inhibitors of apoptosis ( e.g. , caspase inhibitors or deletion of BAX/BAK) or necroptosis ( e.g. , necrostatins or necrosulfonamide) should be examined to rule out these mechanisms as being required for cell death that is postulated to be ferroptosis. Cells dying by ferroptosis primarily exhibit shrunken and damaged mitochondria by electron microscopy, with few other morphological changes evident prior to the point of cell death ( Abrams et al., 2016 ; Dixon et al., 2012 ; Yagoda et al., 2007 ); in particular, nuclei remain intact during ferroptosis ( Dolma et al., 2003 ; Friedmann-Angeli et al, 2014 ). In contrast, apoptosis typically involved fragmentation and margination of chromatin, as well as generation of apoptotic bodies and plasma membrane blebbing. Necroptosis has been demonstrated to involve swelling of cells ( Berghe et al., 2010 ) and was shown to be associated with the generation of broken pieces of plasma membrane that are released from the plasma membrane as vesicles; these “bubbles” stain positive for annexin V as a result of ESCRT-III activation and are released before plasma membrane rupture ( Gong et al., 2017 ). Such features have not been observed during ferroptosis and therefore differentiate these two forms of regulated necrosis. Another form of regulated necrosis is pyroptosis and critically relies on the protein gasdermin D; time-lapse videos of pyroptotic cells demonstrated rapid necrotic cell death that is preceded by intensive blebbing ( Ding et al., 2016 ). This typical morphology has not been seen during ferroptosis. Thus, ferroptosis has unique morphological characteristics. One major outstanding question is why ferroptosis exists, and whether it is adaptive or indicative of a biochemical failure in cells. One possibility is that the incorporation of polyunsaturated fatty acids (PUFAs) into cell membranes during evolution was highly advantageous, allowing more complex physiological functions, including the development of complex neuronal circuits, modulating membrane fluidity, and allowing cells to adapt to environments with different temperatures, among other beneficial functions ( Barelli and Antonny, 2016 ). At the same time, use of PUFAs in membranes creates a new Achilles' heel in the form of susceptibility to lethal lipid peroxidation. These fatty acids possess methylene (CH 2 ) groups flanked on both sides by carbon–carbon double bonds, and one of these hydrogen atoms can be removed to make a highly stabilized pentadienyl radical. The ability to form stabilized radicals makes PUFAs susceptible to peroxidation in the presence of other radical species, as well as metals, oxygen, and specific enzymes. Moreover, oxygenated PUFAs biosynthesized by specialized enzymes – cyclooxygenases, lipoxygenases and cytochromes P450 – are broadly used for signaling purposes. Paradoxically, the signaling benefits of these enzymatic oxygenation reactions are also associated with a risk of generating reactive electrophiles targeting nucleophilic sites in vital proteins. It is possible, that one of the physiological roles of the ferroptotic program is to maximize the benefits and minimize the risk by eliminating cells with excessive production of electrophilic intermediates that cannot be removed. In summary, ferroptosis represents a unique form of regulated cell death, with many of its physiological roles yet to be defined. Using the criteria and reagents described herein may allow the scientific community to determine the physiological and pathophysiological functions of ferroptosis in the years to come.