Ferroptosis: mechanisms, biology, and role in disease

Abstract

The research field of ferroptosis has been enjoying exponential growth over the past few years, since the term was coined in 2012. This unique modality of cell death, driven by iron-dependent phospholipid peroxidation, is regulated by multiple cellular metabolic events, including redox homeostasis, iron handling, mitochondrial activity, and metabolism of amino acids, lipids and sugars, in addition to numerous signaling pathways relevant to disease. Intriguingly, therapy-resistant cancer cells, particularly those of the mesenchymal state and prone to metastasis, are exquisitely vulnerable to ferroptosis. Further, numerous organ injuries and degenerative pathologies are driven by ferroptosis. As such, pharmacological modulation of ferroptosis, via both its induction and inhibition, holds great potential for the treatment of drug-resistant cancers, ischemic organ injuries, and other degenerative diseases linked to overwhelming lipid peroxidation. In this Review, we seek to provide an extensive and critical analysis of the current understanding of the molecular mechanisms and regulatory networks of ferroptosis, the potential physiological functions of ferroptosis in tumor suppression and immune surveillance, and its pathological roles and potential for therapeutics. Importantly, as in all rapidly evolving new research areas, issues and confusions exist due to misconceptions and inappropriate use of experimental tools – this Review also tries to address these issues and to provide practical guidelines. Finally, we discuss important concepts and pressing questions that should be a focus of future ferroptosis research.

Cells represent the fundamental organizing unit of life. Cell proliferation, differentiation, functional characteristics, and ultimately cell death, are therefore of critical importance in the myriad manifestations of life. The fate and functions of cells are impacted by both environmental and genetic cues. One of the most critical axes for cell fate determination is how cells respond to oxidative stress, since most living organisms rely on oxygen as the ultimate electron acceptor in reductive/oxidative (redox)-based metabolic processes. Among the conditions and oxygen-derived species that cause oxidative stress in cells, the oxidative modification of lipids in membrane bilayers (e.g., lipid peroxidation) has emerged as a nexus for integrating a series of environmental and genetic inputs, including heat and radiation exposure, metabolism, redox homeostasis, immune surveillance, and intercellular contacts, as well as oncogenic and tumor suppressive signaling. In fact, these seemingly disparate processes are tightly integrated to instruct cells’ decisions to undergo ferroptosis, a distinct cell death paradigm resulting from unrestrained lipid peroxidation (Fig. 1). Ferroptosis is now appreciated as likely one of the most widespread and ancient forms of cell death, as although originally studied in mammalian systems¹, ferroptosis-like cell death has also been observed in evolutionarily remote species, such as those belonging to the kingdoms of plants, protozoa, and fungi^2–4.

Figure 1. An overview of ferroptosis.

A schematic chart showing that ferroptosis is executed by phospholipid peroxidation, a process relying on metabolic products reactive oxygen species (ROS), phospholipid containing polyunsaturated fatty acid chain(s) (PUFA-PL), and transition metal iron, and that intra- and intercellular signaling events and environmental stresses can impact ferroptosis by regulating cellular metabolism and ROS level. The figure also shows the role of ferroptosis in disease and its potential physiological functions.

Historically, cell death was initially considered to be passive and unregulated, until apoptosis was discovered in the 1970s as the prototypical form of programmed cell death (PCD), as it is executed by a developmentally programmed pathway⁵. The broader category of regulated cell death (RCD) refers to death programs that are molecularly regulated, but not necessarily developmentally programmed⁶. Ferroptosis meets this RCD criterion: it is driven by lethal lipid peroxidation, a consequence of cellular metabolism and imbalanced redox homeostasis, and it can be suppressed by blocking lipid peroxidation directly or through depleting iron, via pharmacological or genetic means. Importantly, mounting evidence suggests potential physiological roles of ferroptosis in tumor suppression and immunity (Fig. 1). More recently, ferroptosis has also been implicated in the development of certain fungal species⁴ and in developmental aging of nematodes⁷. Furthermore, the pathophysiological relevance of ferroptosis, especially as a therapeutic modality in cancer and ischemic organ damage, has been convincingly established (Fig. 1).

Although the term “ferroptosis” was coined fairly recently in 2012, ferroptosis-like cell death was observed long before, for instance, as a type of oxidative-stress-induced cell death termed “oxytosis”⁸. Pioneering work by Harry Eagle in the 1950s and 1960s had shown that deprivation of the amino acid cyst(e)ine can cause cell death⁹, and that the endogenous synthesis of cysteine using methionine and glucose, i.e., transsulfuration, makes cells resistant to such cell death^10,11. Cysteine, either taken up in its oxidized form as cystine by the system x_c⁻ cystine/glutamate antiporter (a transmembrane protein complex containing subunits SLC7A11 and SLC3A2)^12,13 and sodium-dependent neutral amino acid transporter B(0)AT1 (SLC6A19), or in its reduced form cysteine by the neutral amino acid transporter, or generated by endogenous production via transsulfuration, is the rate-limiting substrate for the biosynthesis of reduced glutathione (GSH)¹⁴. GSH, the most abundant reductant in mammalian cells, is important for iron-sulfur cluster biogenesis and is a cofactor for multiple enzymes, including glutathione peroxidases (GPX) and glutathione-S-transferases. It has been demonstrated genetically that GSH synthesis, system x_c⁻, and glutathione peroxidase 4 (GPX4) can all protect cells from death triggered by diverse oxidative stress conditions, particularly those causing thiol deprivation, including inhibition of system x_c⁻ activity^15–19. As the roles of GSH, system x_c⁻, and GPX4 become established in ferroptosis, we can now put all these early studies in the context of ferroptosis.

The field of ferroptosis is nonetheless nascent in many ways, having only coalesced from the adjacent fields of amino acid and lipid metabolism, iron homeostasis, redox and selenium biology, and cell death in the last few years. There has been an exponential growth in the number of published studies on ferroptosis, necessitating a thorough and critical look at recent advances to crystalize for diverse communities of researchers the key findings, questions, and challenges surrounding this topic. Herein, we provide an in-depth analysis of the mechanisms and regulation of ferroptosis, its potential physiological functions, and role in disease and therapy. We discuss emerging questions and challenges that are conceptually important for the field. We also provide practical and experimental suggestions to guide ferroptosis research.

Mechanisms governing ferroptosis

Recent years have witnessed rapid progress in the mechanistic understanding of ferroptosis. Through the initial discovery of the role of the cystine-import-GSH-GPX4 machinery in suppressing ferroptosis, the role of phospholipid hydroperoxides (PLOOHs) as the executioners of ferroptosis is now established. More recently, GPX4-independent ferroptosis surveillance pathways have been identified. Furthermore, the mechanisms of PLOOH synthesis, particularly the synthesis and activation of polyunsaturated fatty acids (PUFAs), the precursor of PLOOHs, have been extensively investigated in the context of ferroptosis. Importantly, all these studies converge on cellular metabolism and have revealed an intimate relationship between ferroptosis and metabolic pathways.

The canonical GPX4-regulated ferroptosis pathway.

In a campaign for the discovery of novel small molecule anti-cancer therapies, the Stockwell Lab conducted a high-throughput screen, beginning in 2001, leading to the discovery of a series of compounds that induce a unique form of non-apoptotic, non-necroptotic cell death reported in 2003²⁰. Counter screening revealed that multiple iron chelators and lipophilic radical-trapping antioxidants (RTAs) inhibited this type of cell death²¹. The iron requirement inspired coining the term “ferroptosis”¹. Mechanistic investigations identified two cellular components, inhibition of which induces ferroptotic death: system x_c⁻ cystine/glutamate antiporter and GPX4, inhibited by the compounds erastin and RSL3, respectively^1,22,23.

GPX4, a selenoprotein originally discovered by Ursini and colleagues through biochemical purification, is the major enzyme catalyzing the reduction of PLOOHs in mammalian cells^24,25. Reduction of phospholipid and cholesterol hydroperoxides to their corresponding alcohols by GPX4 requires the catalytic selenocysteine residue of GPX4 and two electrons provided mainly by GSH, but also by other low molecular thiols or even protein thiols²⁶. The thorough study of the first conditional Gpx4 knockout mouse model by the Conrad group provided early evidence that loss of Gpx4 causes lipid-peroxidation-dependent, non-apoptotic cell death in murine embryonic fibroblasts, and neurodegeneration in hippocampus and cortical regions of the brain¹⁹. This and several other mouse models have helped to delineate a more detailed picture of the in vivo relevance of ferroptosis as discussed further later.

As GPX4 is the major PLOOH-neutralizing enzyme, a general mechanism underlying erastin/RSL3-induced ferroptosis emerged: both compounds inactivate GPX4 – RSL3 does so directly, and erastin does so indirectly by inhibiting cystine import, thus depriving cells of cysteine, an essential cellular antioxidant and a building block of GSH. Consequently, PLOOHs accumulate, possibly causing rapid and unrepairable damage of plasma membrane, leading to cell death (Fig. 2A). Conceptually, these findings establish ferroptosis as a cell death modality with mechanisms distinct from other known death processes. The pharmacological and genetic tools developed herein enable, and have become indispensable for, ferroptosis research.

Figure 2. Ferroptosis-suppressing pathways.

(A) The canonical ferroptosis controlling axis entails uptake of cystine via the cystine-glutamate antiporter, designated system x_c⁻, glutathione (GSH)- and/or thioredoxin reductase 1 (TXNRD1)-dependent reduction of cystine to cysteine, GSH biosynthesis, and glutathione peroxidase 4 (GPX4)-mediated reduction of phospholipid hydroperoxides (PL-OOH) yielding the corresponding alcohols (P-LOH). Recycling of oxidized glutathione (GSSG) is achieved via glutathione-disulfide reductase (GSR) using electrons provided by NADPH/H⁺. (B) In two independent genetics screens the FSP1/ubiquinone (CoQ₁₀) system has been recently identified that completely protects against ferroptosis induced by pharmacological inhibition or genetic deletion of GPX4. FSP1 prevents lipid peroxidation and associated ferroptosis via reduction of ubiquinol/α-tocopherol on the level of lipid radicals unlike GPX4/GSH. (C) Alternate ferroptosis-suppressive mechanisms include squalene- and di-/tetrahydrobiopterin (BH₂/BH₄)-mediated inhibition of lipid peroxidation, although the chemical mechanisms how this is achieved remains to be shown (Abbreviations: ACSL4, acyl-CoA synthetase long chain family member 4; FDFT1, Farnesyl-diphosphate farnesyltransferase 1; GCH1, GTP cyclohydrolase 1, LOX, lipoxygenase; LPCAT3 lysophosphatidylcholine acyltransferase 3, POR, cytochrome P450 oxidoreductase, PUFA, polyunsaturated fatty acid).

Phospholipid peroxidation.

Unrestrained lipid peroxidation is the hallmark of ferroptosis. Early research in the 1950s pointed towards a nexus for the suppression of lipid peroxidation by the trace element selenium, as well as vitamin E and cysteine^27,28. Initiation of lipid peroxidation requires the abstraction of a bisallylic hydrogen atom, in between two carbon-carbon double bonds, from polyunsaturated fatty acyl moieties present on phospholipids (PUFA-PLs) incorporated in lipid bilayers, forming a carbon-centered radical, and subsequent reaction with molecular oxygen to yield a peroxyl radical^29,30 (Fig. 3). If not converted to a lipid hydroperoxide and reduced to the corresponding alcohol, propagation of free-radical-mediated reactions lead to formation of a myriad of secondary products, breakdown of membrane integrity and ultimately rupturing of organelle and cell membranes. Thus, membranes containing a high PUFA-PL content, as evident in neurons, are particularly vulnerable to peroxidation. In line with this, genome-wide haploid and CRISPR/Cas9-based screens unveiled two membrane-remodeling enzymes, acyl-CoA synthetase long chain family member 4 (ACSL4) and lysophosphatidylcholine acyltransferase 3 (LPCAT3)^31–33, as important drivers of ferroptosis. The role of ACSL4 in the ferroptotic process is based on its ability to ligate preferably long chain PUFAs, including arachidonic acid (20:4) and adrenic acid (22:4), with coenzyme A, whereupon they can be re-esterified in phospholipids by various LPCAT enzymes. Genetic loss of ACSL4 causes a dramatic shift from long chain PUFA tails to short chain and monounsaturated fatty acyl tails (MUFAs) in phospholipids^31,34, in a manner similar to what is seen in wild-type cells treated with pharmacological inhibition of ACSL4³⁵. Such a dramatic change in the phospholipidome of ACSL4-null cells enables proliferation of Gpx4 knockout cells for months, a genetic rescue of Gpx4 that was not observed before³¹. Along the same line, exogenous supplementation of MUFAs, stearoyl-CoA desaturase (SCD1)-mediated cellular MUFA production, and an ACSL3-dependent enrichment of membranes with MUFAs was reported to lower the cells’ propensity to succumb to ferroptosis^36–38. Notably, expression of ACSL4 in a subset of triple negative breast cancer cell lines correlates with their sensitivity towards ferroptosis inducers³¹, a correlation that seems to be shared with therapy-resistant, mesenchymal cancer cells³⁹ and clear-cell renal carcinoma³³. Suppression of ACSL4 expression thus may be a principal mechanism in desensitizing cells to ferroptosis, regulated by diverse signaling pathways^40,41. Conversely, increased expression/activity of ACSL4 might promote ferroptosis under various pathophysiological contexts^42,43.

Figure 3. Mechanisms of phospholipid peroxidation.

Lipid peroxidation, the hallmark of ferroptosis, occurs in both non-enzymatic and enzymatic ways (marked by a dashed box). For the latter, lipoxygenases (LOXs) and/or cytochrome P450 oxidoreductase (POR) have been implied in initiating the process of lipid peroxidation by dioxygenation of lipids, although definitive genetic evidence for an involvement of lipoxygenases in the ferroptotic process is lacking. Lipid peroxidation can be divided in three phases, i.e. initiation, propagation and termination, as indicated with the differently colored arrows. The lipid peroxidation-inhibiting systems – involving enzymes and small molecules – act on the different levels of the lipid peroxidation cascade (Abbreviations: ETC, electron transport chain; FSP1, ferroptosis suppressor protein 1; Fe²⁺, ferrous iron, Fe³⁺, ferric iron, GPX4, glutathione peroxidase; LOX, lipoxygenase; POR, cytochrome P450 oxidoreductase; L•, lipid radical; L-H, lipid; L-O•, alkoxyl radical; L-OO•, peroxyl radical; L-OH, lipid alcohol; L=O, lipid carbonyl, NOX, NADPH oxidase; OH⁻ hydroxide ion; O₂·⁻, superoxide anion; RTA, radical trapping antioxidant).

While there is no doubt that the degree of unsaturation of lipid bilayers is key in determining the sensitivity of cells to ferroptosis, there are many uncertainties and debates on how lipid peroxidation is initiated. A bis-allylic carbon adjacent to a conjugated diene has among the weakest of C–H bonds known and an increase in the number of these structures augments the rate of autoxidation of lipids, as is evident in the spoiling of PUFA-containing foods under ambient oxygen tension. Conceivably, non-enzymatic initiation of lipid peroxidation can be sparked by alkoxyl radicals in lipids or hydroxyl radicals, in a Fenton-type radical chemistry using iron as the catalyst⁴⁴ (Fig. 3) (detailed later).

Certain lipoxygenases (LOXs), which are non-heme iron-dependent dioxygenases, can directly oxygenate PUFAs of biological membranes⁴⁵, raising the prospect that LOXs may mediate ferroptosis. This possibility was supported by the observations that some pharmacological inhibitors of LOX can inhibit ferroptosis^19,46, and that 12/15-LOX knockout or LOX inhibitor baicalein protected mice against ischemic brain injury and edema formation^47,48. Nonetheless, genetic removal of 12/15-LOX on the Gpx4 knockout background repeatedly failed to prevent ferroptosis in mouse fibroblasts, acute kidney injury and associated lethality in vivo⁴⁹, nor did it restore CD8⁺ T cells in T-cell specific Gpx4^−/− mice⁵⁰. These data suggest that alternative mechanisms compensate for Alox15 loss and/or that the frequently used “LOX-specific” inhibitors exert non-specific RTA activity⁵¹. Indeed, a recent study confirmed that most frequently used LOX inhibitors harbor RTA activity⁵², thus challenging a substantial role of LOXs in ferroptosis. Further, combined downregulation of all human LOX isoenzymes failed to prevent RSL3-induced ferroptosis, although it provided substantial rescue against erastin-induced ferroptosis, likely due to the fact that erastin treatment activates lipoxygenases³⁷.

Can individual LOXs play a more prominent role in ferroptosis, at least under certain specific contexts? Deletion of Alox15 or Alox12 showed protective roles in specific mouse models of neurodegeneration or cancer suppression^47,53, respectively. However, it should be stressed that these enzymes play important physiological roles in the immune system by directly modulating (neuro)inflammatory processes and the tumor microenvironment via generating pro- and anti-inflammatory molecules^54,55, thus not necessarily directly related to ferroptosis. Equally relevant to this notion are previous findings that GPX4 controls the activities of lipoxygenases and cyclooxygenases by the so-called cellular peroxide tone, as both types of eicosanoid-metabolizing enzymes require oxidation of their iron by lipid hydroperoxide to capture and incorporate molecular oxygen into PUFAs⁵⁴. Notably, it has been reported that ALOX12 is essential for p53-dependent ferroptosis triggered by peroxides, and this form of ferroptosis is rather unique as it appears to be independent of ACSL4⁵³; PE-binding protein-1 (PEBP1) has been reported to complex with certain LOXs, altering their substrate specificity towards PUFA-PLs⁵⁶.

Besides the controversial role of LOXs in ferroptosis, recent findings suggest that the ubiquitously expressed cytochrome P450 oxidoreductase (POR) plays a role in initiating lipid peroxidation⁵⁷. After accepting electrons from POR by using NADPH as electron donor, downstream electron acceptors, such as cytochrome P450 and CYB5A, are reduced, which could subsequently either directly or indirectly trigger lipid peroxidation by abstracting methylene hydrogen from PUFAs or by reducing ferric iron^57,58.

Despite the widespread lack of PUFAs in bacterial membranes, Pseudomonas aeruginosa expresses a secreted lipoxygenase (PA-LOX)⁵⁹. This enzyme was shown to induce oxidation of membrane lipids of human red blood cells⁶⁰, and was able to induce ferroptosis in human bronchial epithelial cells⁶¹. Accordingly, patients suffering from cystic fibrosis presented increased levels of oxidized arachidonic acid in phosphatidylethanolamine⁶¹, which was reported to be one of the main phospholipid species detected in cells and tissues dying by ferroptosis³⁴. Such a ferroptosis-inducing mechanism across organisms is intriguing and warrants further investigations to determine whether this mechanism is exploited by other lower organisms.

Cell metabolism and ferroptosis.

It is obvious that multiple metabolic reactions can lead to the cellular production of PLOOHs, and mounting evidence has demonstrated that metabolism plays a central role in ferroptosis. Studies by Jiang and colleagues, seeking to determine how metabolism contributes to cell fate determination unraveled the intricate relationship of ferroptosis with metabolism^62–64. They started by chasing the elusive cell-death-promoting function of autophagy. As a crucial survival mechanism in response to various stresses, whether and how autophagy can also promote cell death (i.e., “autophagic cell death”) has been debated for decades⁶⁵. Unexpectedly, they found that autophagy promotes a rapid non-apoptotic, non-necroptotic form of cell death, upon amino acid starvation (a condition that triggers potent autophagy), but only in the presence of full serum. They subsequently found that the iron-carrier transferrin and the amino acid glutamine in serum were required for this form of cell death and that deprivation of the amino acid cystine from the cell culture medium was sufficient to trigger the death. Hence, this type of cell death is ferroptosis induced by cystine starvation (mimicking erastin) and dependent on iron-transferrin⁶². The role of autophagy in cystine-deprivation-induced ferroptosis is via autophagic degradation of the iron-storage protein ferritin (aka ferritinophagy), which results in an increase of cellular labile iron content, and thus sensitization to ferroptosis^63,66 (Fig. 4). The requirement of glutamine metabolism, or glutaminolysis, for cysteine-deprivation-induced ferroptosis⁶² directly links ferroptosis to cell metabolism. The anaplerotic function of glutaminolysis in generating α-ketoglutarate to fuel the mitochondrial TCA cycle suggests the involvement of normal metabolic function of mitochondria in ferroptosis, which was subsequently validated through a variety of pharmacological, cellular, and genetic analyses⁶⁴ (Fig. 4). Notably, mitochondria have been shown earlier to be active participants in oxytosis^8,67.

Figure 4. Metabolism and cell signaling in ferroptosis.

The figure depicts the regulation of ferroptosis by multiple metabolic events (such as lipogenesis, autophagy, and mitochondrial TCA cycle) and signaling pathways (such as E-cadherin-NF2-Hippo-YAP pathway, glucose-regulated AMPK signaling, and p53 and BAP1 tumor suppressor function). See text for detail. Abbreviations: TfR, transferrin receptor; PL-OOH, phospholipid hydroperoxide; PUFA-PL, phospholipid with polyunsaturated fatty acid chain; ROS, reactive oxygen species; TCA, mitochondrial TCA cycle; Gln, glutamine; Glu, glutamate; αKG, α-ketoglutarate.

The involvement of mitochondria in ferroptosis underscores its metabolic nature (Fig. 4), and based on these findings, one can also raise a plethora of hypotheses and questions. For example, would glucose, a major fuel of mitochondrial TCA cycle, regulate ferroptosis? Indeed, glucose starvation has been recently shown to suppress ferroptosis^68,69. However, mechanistically this appears to be mainly due to AMPK signaling, instead of TCA cycle engagement (detailed later). And on a related note, if the ferroptotic function of glutamine is solely through its anaplerotic role in the TCA cycle, then how can glutamine be essential for cysteine-deprivation-induced ferroptosis even when cells are cultured in the presence of abundant glucose⁶²? Thus, a detailed study of both TCA-cycle-dependent and independent roles of glutaminolysis in ferroptosis is warranted. Furthermore, and fundamental to ferroptosis, does mitochondrial metabolism promote ferroptosis by generating specific lipid precursors for PLOOH synthesis (intermediates of the TCA cycle can participate in lipogenesis), or via the generation of reactive oxygen species (ROS; natural byproducts of oxidative metabolic reactions)? The observation that both the mitochondrial activity and glutaminolysis are crucial for cysteine-deprivation-induced ferroptosis, but are dispensable for that induced by GPX4 inhibition⁶⁴, lend more support to the latter mechanism.

Besides mitochondria, plant cells possess another unique organelle specialized in metabolism – the chloroplast – to conduct photosynthesis, which comprises a chain of oxidative-reductive anabolic reactions. As ferroptosis has been observed in plants^2,70–72, an intriguing speculation is that chloroplasts may also play an important role in the regulation of ferroptosis in plants.

Iron in ferroptosis.

The essential function of cell metabolism, especially phospholipid peroxidation, in ferroptosis provides insights into why this death modality depends on iron. First, LOXs and POR, metabolic enzymes implicated in phospholipid peroxidation, require iron for catalysis; and iron is also essential for a plethora of metabolic enzymes involved in the generation of cellular ROS. Second, the non-enzymatic, iron-dependent Fenton chain reaction is likely essential for ferroptosis: when GPX4 is inhibited, PLOOHs can persist longer, initiating Fenton reaction to rapidly amplify PLOOHs, the hallmark of ferroptosis²⁹ (Fig. 3). PLOOHs can react with both ferrous and ferric ions to generate free radicals PLO• and PLOO•, respectively, and these free radicals react with PUFA-PLs to further propagate PLOOH production.

Given the central role of iron in cell viability and death, it is not a surprise that cellular iron homeostasis is under exquisite control, mainly through a posttranscriptional network dictated by iron-regulatory proteins IRP1 and IRP2^73,74 to regulate intracellular iron storage/release and import/export. Conceivably, many cellular processes alter the sensitivity of cells toward ferroptosis by changing cellular labile iron contents. Transferrin and its receptor work together to promote ferroptosis by importing iron into the cell⁶². Conversely, mechanisms that enhance cellular iron export have been shown to render cells more resistant to ferroptosis^75,76. As discussed, ferritin and its autophagic degradation can also regulate ferroptosis^63,66. Further, liberating iron from degrading heme via hemeoxygenase-1 (HO-1)-mediated heme degradation has also been implicated in ferroptosis; yet a series of conflicting data suggest HO-1 to either promote or suppress ferroptosis^77,78.

Recent in vivo mouse model studies further illustrated the role of iron regulation in ferroptosis. For example, genetic deletion of ferritin heavy chain promotes cardiomyopathy, likely via enhancing ferroptosis⁷⁹. Intriguingly, hepatocyte-specific knockout of the transferrin gene in mice led to rather unexpected phenotype: feeding the knockout mice with high dietary diet increased iron loading in hepatocytes, cells normally responsible for transferrin synthesis in the body, and rendered the mice more susceptible to liver fibrosis, which could be ameliorated by lipophilic RTAs⁸⁰. This study suggests that in the absence of transferrin, SLC39A14 is compensatorily upregulated in hepatocytes, leading to excessive import of iron.

GPX4-independent surveillance pathways.

Although the cyst(e)ine-GSH-GPX4 axis is considered the main ferroptosis-regulating system in mammals^22,49, genome-wide screens have recently uncovered GPX4-independent mechanisms. Using either a genetic suppressor screen or a synthetic lethal CRISPR-Cas9 screen, the Conrad group and Olzmann group independently identified ferroptosis suppressor protein 1 (FSP1) as the second mainstay of ferroptosis^81,82, acting differently from the thiol-dependent axis. FSP1 was previously called AIFM2^83,84, as it is homologous with AIFM1 (apoptosis inducing factor mitochondria associated 1)⁸⁵. AIFM1, initially considered to be pro-apoptotic (like AIFM2), is nowadays considered to mediate transport and proper folding of mitochondrial intermembrane proteins instead⁸⁶. Similarly, FSP1 lacks substantial pro-apoptotic function, but rather protects cells from ferroptosis induced by inhibition or genetic deletion of GPX4^81,82.

FSP1 is myristoylated and associates with several cell membrane structures including plasma membrane, Golgi apparatus, and perinuclear structures. Mutation of the myristoylation site abrogates its cell-protective function. Mechanistically, due to its NADH:ubiquinone oxidoreductase activity⁸⁷, FSP1 suppresses lipid peroxidation and ferroptosis by either reducing ubiquinone (CoQ₁₀)/semihydroquinone to yield ubiquinol, which in turn may directly reduce lipid radicals to terminate lipid autoxidation, or indirectly via regenerating oxidized α-tocopheryl radical (vitamin E)^81,82, the most powerful natural chain breaking antioxidant in lipids (Fig. 2B). These studies thus also solved a long-standing mystery why there is so much extramitochondrial ubiquinone in some cells and tissues⁸⁸, in contrast to its canonical role in the mitochondrial electron transport chain.

Using a CRISPR/Cas9-based activator screen, GTP cyclohydrolase-1 (GCH1) was recently reported to protect against ferroptosis via its metabolic products tetrahydrobiopterin (BH₄) and dihydrobiopterin (BH₂)⁸⁹. BH₄ was shown to prevent depletion of phospholipids containing two polyunsaturated fatty acyl tails, likely involving a dual mechanism by both acting as a direct RTA and being involved in CoQ₁₀ synthesis^89,90 (Fig. 2C). While the role of GCH1 in protecting tissues and organs from ferroptosis remains to be elucidated, knockout studies showed that loss of Gch1 in mice causes bradycardia and embryonic death during mid-gestation⁹¹. Besides these systems which either directly act on peroxides in lipid bilayers or on phospholipid radicals via naturally occurring RTAs, still other cell-intrinsic mechanisms may exist that protect against deleterious lipid peroxidation. In this context, accumulation of squalene, a metabolite of the cholesterol pathway, was reported to confer anti-ferroptotic activity in cholesterol-auxotrophic ALK (anaplastic lymphoma kinase)-positive anaplastic large cell lymphoma cell lines and primary tumors, albeit it remains to be shown whether this is a cancer subtype-specific effect or a general protective mechanism⁹² (Fig. 2C).

Regulation of ferroptosis

Conceivably, biological processes that modulate ferroptosis-promoting or surveillance molecules, redox and iron homeostasis, and cell metabolism could impact ferroptosis. As expected, the oxidative-stress-responsive transcription factor NRF2 can mitigate ferroptosis by stimulating the expression of its multiple canonical target genes (see⁹³ for a comprehensive review). Additionally, mounting evidence has demonstrated that under specific biological contexts, multiple signaling pathways can dictate the susceptibility of cells to ferroptosis.

Regulation of GPX4 and FSP1.

Despite their prevailing importance for ferroptosis, little is known how GPX4 and FSP1 are regulated on the transcriptional, translational and, equally important, post-translational/activity level, under both physiological and pathological conditions. Nevertheless, it seems that both GPX4 and FSP1, at least to some extent, intersect with the mevalonate pathway: isopentenylation stabilizes the selenocysteine-specific tRNA (Trsp), which is required for the synthesis of selenoenzymes including GPX4, and as one of the final metabolites of the mevalonate pathway, ubiquinone (CoQ10), is a major substrate of FSP1⁹⁴.

As one of 25 dedicated selenoproteins in humans, GPX4 expression is tightly regulated⁹⁴. For instance, selenium supplementation was shown to boost its expression through an SP1- and-TFAP2c-dependent manner in neurons⁹⁵. However, it should be noted that GPX4 can be regarded a housekeeping protein being constitutively expressed in most tissues and organs, unlike those true selenium-responsive proteins such as selenoprotein P (SELENOP), GPX1, and GPX3. Other transcription factors that have been reported to regulate GPX4 expression include CEBP1 and C/EBPs in enterocytes⁹⁶, and NF-Y in some cancer cells⁹⁷. Post-transcriptionally, guanine-rich sequence-binding factor 1 (GRSF1) was reported to bind to the 5’UTR of the mitochondrial form of the Gpx4 mRNA, causing increased translation of mitochondrial GPX4 that plays an essential role during sperm development^98,99. The relevance of these regulatory events to ferroptosis, however, remains undefined.

Accumulating evidence suggests that GPX4 is regulated on the activity and stability level. For instance, impaired GSH-dependent reduction of the active site selenocysteine, due to sustained oxidative stress and concomitant lack of GSH may cause irreversible inactivation of GPX4 by formation of a redox-dead dehydroalanine in a process known as ß-cleavage¹⁰⁰. Moreover, formation of a selenylamide between the selenenic acid and a neighboring amino acid may protect the enzyme from irreversible inactivation. However, it needs to be further explored whether any of these mechanisms are at play during pathological conditions, such as ischemia/reperfusion injuries (IRI), albeit it has been reported that intestinal IRI is associated with much lower GPX4 levels⁴². In addition, several well-established ferroptosis inducers including RSL3 ultimately cause GPX4 depletion, either through covalent inhibition of the active site selenocysteine, impairment of the mevalonate metabolism or general iron-dependent oxidative stress^101–103.

As another prominent ferroptosis suppressor, FSP1 was originally described as a p53-responsive gene, therefore initially called p53-responsive gene 3 (PRG3)¹⁰⁴. FSP1 is a target of transcription factors NRF2¹⁰⁵, CEBP¹⁰⁶, and PPARα¹⁰⁷. Interestingly, in T-cell lymphoblastic lymphoma cells, FSP1 expression was reported to be upregulated by the long non-coding RNA (lncRNAs) maternally expressed 3 (MEG3) gene and suppressed by miR-214¹⁰⁸, two regulatory RNAs involved in tumor development. Beyond transcriptional regulation, almost nothing is known how the oxidoreductase activity of FSP1 is regulated and how its differing subcellular localization impacts on its role in different physiological and pathophysiological processes^81–83,106. But the promiscuity of FSP1 toward both the reducing and oxidizing substrates (including NADH, NADPH, CoQ10 and α-tocopherol) suggests its complex regulation.

E-cadherin-NF2-Hippo-YAP signaling and ferroptosis.

The Hippo-YAP pathway is involved in a myriad of biological functions, including cell proliferation and organ size control^109,110. The investigation of the role of this pathway in ferroptosis was initiated from an interesting observation that cells grown with high density are often more resistant to ferroptosis induced by both cysteine deprivation and GPX4 inhibition⁴¹. This observation is reminiscent of earlier observations; a directly relevant one is that cells were shown to survive Gpx4 knockout if they are cultured at extremely high density or as spheres^19,111. Mechanistically, the cell density effect on ferroptosis in epithelial cells is mediated by E-cadherin-mediated cell-cell contacts, and such intercellular interaction activates intracellular Hippo signaling through NF2/merlin tumor suppressor, and thus inhibits the transcription co-regulatory activity of oncoprotein YAP; various target genes of YAP, including ACSL4, TfR1, and possibly others, are regulators of ferroptosis⁴¹ (Fig. 4). Consistent with this finding, TAZ, the close homolog of YAP, has also been shown to enhance ferroptosis in a cell-density-regulated manner in kidney cancer cells that predominantly express TAZ instead of YAP¹¹². Further, a closer analysis of the target gene profile of YAP and TAZ indicates that YAP and TAZ transcriptionally modulate various processes relevant to ferroptosis, including iron homeostasis, oxidative response, and cellular metabolism^113,114.

The role of the E-cadherin-NF2-Hippo-YAP pathway in dictating ferroptosis sensitivity has important implications. First, as multiple components of this pathway are frequently mutated in cancer, ferroptosis induction might be exploited as a potential therapeutic approach for the treatment of these specific cancers, a topic further discussed later. Second, as such cell density-dependent ferroptosis was also observed in non-epithelial cells that do not express E cadherin⁴¹, would other cadherins or cell adhesion molecules also mediate similar mechanisms for ferroptosis suppression? Thirdly, the Hippo-YAP pathway is important in development and interacts with various other signaling pathways, thus one might investigate the potential function of ferroptosis in normal biology through its link to the Hippo-YAP pathway. Speculatively, did cadherins evolve in multicellular organisms in such a way that they not only physically connect cells, mediate intercellular communication via transducing signals into intracellular machinery, but also function as an ancient mechanism to protect cells from oxidative stress challenges and its most devastating consequence, ferroptosis? All these events are essential for the being of a multicellular organism. Indeed, the expression of cadherin proteins can be traced back all the way to some ancestral metazoan species¹¹⁵.

AMPK signaling and ferroptosis.

Intuitively, energy and metabolic stress should result in the loss of energy and thus a cascading failure of systems needed to maintain homeostasis, such as energy-dependent ion gradients across cell membranes¹¹⁶, ultimately resulting in cell death. Moreover, metabolically stressing cells with glucose starvation increases ROS production¹¹⁷, suggesting that glucose starvation promotes ferroptosis. Surprisingly, glucose starvation instead blocks ferroptosis^68,69. This protective effect was dependent on the activity of energy-sensing kinase, AMPK. Hence, when glucose is absent, AMPK is activated, turning on an energy stress-protective program against ferroptosis that involves impaired biosynthesis of PUFAs, which are essential for lipid peroxidation-driven ferroptosis^34,37 (Fig. 4). This finding has practical applications, as activating this energy stress program was found to protect against renal IRI⁶⁸. It may be that this protective mechanism exists as a first line of defense against organ injury, which can often involve energy failure along with the injury.

Hypoxia signaling and ferroptosis.

Given that ferroptosis is death by phospholipid peroxidation, a longstanding question has been whether it is dependent on oxygen concentrations. An early experiment indicated little loss of sensitivity to erastin-induced ferroptosis in 1% oxygen, suggesting that hypoxia does not suppress ferroptosis²³. A recent study indicated that hypoxia sensing can in fact drive sensitivity to ferroptosis: clear cell carcinomas were found to be highly sensitive to ferroptosis induced by GPX4 inhibition, and this sensitivity was driven by HIF2α through PUFA lipid remodeling via hypoxia-inducible, lipid droplet-associated protein (HILPDA)³³. Clear cell carcinomas have a characteristic clear cytoplasmic staining upon histology staining and are difficult to treat¹¹⁸, indicating clinical implication of this finding. Also, it is tempting to speculate that the HIF2α-HILPDA-driven sensitization to ferroptosis may be an ancient means to eliminate nascent hypoxic tumors.

Potential biological functions of ferroptosis

The mechanisms of ferroptosis suggest its potential evolutionary origin: ever since living organisms on Earth began to use oxygen to drive metabolism, essentially through a series of chemical redox reactions, the transition metal iron has become the prime factor to catalyze these reactions. An inevitable side product of such iron-dependent, redox-based metabolism are ROS, including PLOOHs. When the cellular level of PLOOHs exceeds a certain threshold, cells succumb to ferroptosis. Conceivably, multiple surveillance mechanisms like GPX4 and FSP1 have evolved to protect cells from ferroptosis. Thus speculatively, ferroptosis might be a most ancient form of regulated cell death.

But is there any beneficial, physiologically relevant role for ferroptosis? In theory, this is possible. A related example might be ROS and lipid hydroperoxides: first considered as mere toxic “byproducts” of metabolism and later found to confer essential physiological functions, for instance in cell signaling and immunity, they are even deliberately generated at several subcellular sites to regulate cellular physiological events^119–121. Similarly, ferroptosis might also be adapted to accomplish roles beneficial for life. Mounting evidence, although indirect, suggests the physiological function of ferroptosis in tumor suppression and immune surveillance.

Ferroptosis in tumor suppression.

Multiple tumor suppressors have been shown to sensitize cells to ferroptosis. Therefore, a reasonable hypothesis is that ferroptosis contributes to the antitumor activity of these tumor suppressors, i.e., tumor suppression might be an innate physiological function of ferroptosis.

Among such tumor suppressors, the involvement of p53 in ferroptosis has been thoroughly investigated. Through a detailed analysis of specific lysine acetylation sites of p53, Gu and colleagues found that p53 can potentiate ferroptosis via suppressing the transcription of system x_c^– subunit SLC7A11, and this function may contribute to the tumor suppressive function of p53 in vitro and in vivo^122,123. Furthermore, a cancer-prone single nucleotide polymorphism of p53, P47S, was found to be enriched in the African population and to make cancer cells more resistant to ferroptosis¹²⁴. While these findings are consistent with a role of ferroptosis in p53-mediated tumor suppression, it is not clear whether the loss of ferroptosis-promoting activity of p53 is the only functional consequence caused these specific mutations. Notably, in contrast to these studies, p53 has also been reported to prevent ferroptosis via modulating its other transcriptional targets^125,126. Given that p53 can regulate a large cohort of target genes that are involved in various biological processes, its precise role in ferroptosis can well be context-dependent.

Similar to p53, the tumor suppressor and epigenetic regulator BAP1 can also promote ferroptosis by downregulating SLC7A11 expression¹²⁷. But unlike p53, whose ferroptosis-promoting activity alone has been suggested to be sufficient to suppress tumorigenesis in vivo¹²³, it is not clear how significantly the ferroptosis-promoting activity of BAP1 contributes to its tumor-suppressive function.

Fumarase, an enzyme catalyzing the conversion of fumarate to malate in the TCA cycle, is a bona fide tumor suppressor in leiomyoma and papillary renal cell carcinoma¹²⁸. It has been elusive how this metabolic enzyme can counter-intuitively suppress tumorigenesis, and how it does so only in a few specific cancer types. The involvement of fumarase in the ferroptosis-promoting function of the mitochondrial TCA cycle⁶⁴ sheds light to these questions: during the development of certain types of cancer, tumor cells encounter substantial oxidative stress and are primed to undergoing ferroptosis. Under these conditions, loss of fumarase function, although impairing growth capability, renders cells more resistant to ferroptosis, thus enhancing their survival and potential of cancerous transformation.

Ferroptosis in immune surveillance.

Unlike other forms of cell death such as apoptosis, pyroptosis and necroptosis, which are clearly engaged by the immune system¹²⁹, it remains largely obscure whether sensitizing or triggering ferroptosis by extrinsic or intrinsic mechanisms may fulfill a similar “physiological role”. Due to the multifaceted nature and the involvement of numerous metabolic pathways directly impinging on ferroptosis, it is highly likely that certain key nodes of the ferroptosis process are targeted by the immune system. Early findings by Sato’s laboratory suggested that system x_c^– activity is suppressed by interferon-γ (IFN-γ)¹³⁰. Such a mechanism is employed by CD8⁺ T cells in sensitizing tumor cells toward ferroptosis¹³¹. Whether such a direct mechanism of immune cells on other ferroptosis nodes might be of physiological relevance remains elusive. Early studies indicated that interleukins 4 and 13 suppress GPX4 expression in certain cells coinciding with an increased expression of ALOX15, thus permitting a full-blown immune activation by metabolites of the arachidonate pathway¹³². Since GPX4 is known to suppress the activities of lipoxygenases and cyclooxygenases by lowering the lipid peroxide tone⁵⁴, it is possible that impaired activity of GPX4 may have a profound effect on the secretion of pro- and anti-inflammatory lipid mediators, which in turn may alert the innate immune system for the presence of cells and tissues in a ferroptosis-sensitive state.

The potential immune function of ferroptosis is not limited to the mammalian system. A recent report suggests that ferroptosis promotes host immunity of rice against infection by fungus M. Oryzae⁴. A ferroptosis-like cell death of specific rice blast cells, associated with lipid peroxidation and inhibitable by iron chelators, prevents M. Oryzae infection. Intriguingly, ferroptosis-like death of specific cells of M. Oryzae is conversely required for its host invasion and development in the host. As such, this finding also implicates, for the first time, a potential developmental role of ferroptosis.

Role of ferroptosis in disease and therapeutic opportunities

Although the physiological function of ferroptosis remains obscure, its role in a plethora of human diseases has been extensively documented. Importantly, pharmacological modulation of ferroptosis has been demonstrated to be a promising therapeutic venue for the treatment of cancer and IRI in diverse preclinical animal models. While in this section we focus on the role and therapeutic implication of ferroptosis in cancer and IRI, ferroptosis has been implicated in the pathogenesis of many other diseases, as discussed in Box-1.

Box 1 – Diseases possibly related to ferroptosis.

In addition to cancer and ischemic organ injuries, ferroptosis has been implicated in the pathogenesis of a growing list of other diseases, such as neurodegeneration^{19,165–167,182}, liver and lung fibrosis^80,183, autoimmune diseases^184,185, Mycobacterium-tuberculosis-induced tissue necrosis¹⁸⁶, cigarette-smoking-associated chronic obstructive pulmonary disease^187,188, and a rare genetic neurological disorder called Pelizaeus-Merzbacher Disease¹⁸⁹. While this long list speaks to the clinical relevance and therapeutic potential of ferroptosis-modulating approaches, further investigation is required to determine if there is indeed a causative role of ferroptosis in these diseases. For example, in most cases the general observation is that non-apoptotic cell death was observed in the disease tissue, and that a ferroptosis inhibitor, often a lipophilic RTA, could mitigate the observed cell death and in some cases, alter the severity of the symptom. However, lipid peroxides can regulate immunity and inflammation, processes that play important roles in all the listed diseases. Therefore, caution is needed to distinguish whether the observed effect of lipophilic RTA is via modulation of inflammation or ferroptosis, or both. A detailed mechanistic interrogation of tissue cell death associated with the disease, including examining specific in vivo ferroptosis biomarkers (which the field sorely miss), will be crucial for this purpose.

Ferroptosis in cancer.

Ferroptosis has been linked to cancer since the very beginning of the field: the initial discovery of chemical inducers of ferroptosis is the result of a hunting for novel cancer therapeutic compounds^20,21. Subsequent mechanistic studies have revealed that numerous cancer-relevant genes and signaling pathways regulate ferroptosis. The observations that mesenchymal and dedifferentiated cancer cells, which are often resistant to apoptosis and common therapeutics, as well as so called “therapy-persister” cancer cells, are highly susceptible to ferroptosis inducers^39,133,134, further underscores the promise of ferroptosis induction as a novel cancer therapy.

Conceptually, since ferroptosis is an oxidative stress-induced, metabolic form of cell death, it seems logical to propose that cancer cells may have higher tendency to undergo ferroptosis, due to overall more active metabolism and higher ROS load. Adding to this notion, it has been shown that cancer cells often demand high iron contents^135,136, which may further sensitize them to ferroptosis. However, cancer cells may also harness additional genetic or epigenetic alterations to counter these metabolic and oxidative burdens, such as increased expression of SLC7A11 or upregulation of the anti-oxidative transcription factor NRF2¹³⁷. Therefore, whether a given cancer is more sensitive or resistant to ferroptosis induction is dictated by its specific genetic background. The genomics of cancer, as well as various other parameters as discussed below, should be considered for the development of ferroptosis induction-based cancer therapy.

Ferroptosis-related cancer biomarkers and genetic alterations

Multiple oncoproteins, tumor suppressors, and oncogenic signal transduction pathways can regulate ferroptosis. Therefore, their alterations in cancer can be used as biomarkers to predict the responsiveness of cancer cells to ferroptosis-inducing therapies.

Taking the E-cadherin-NF2-Hippo-YAP pathway as an example and based on the TCGA analysis, loss of function mutation of tumor suppressor E cadherin is a frequent event in breast lobular invasive carcinoma (~65%) and diffusive gastric adenocarcinoma (~25%), among other cancers; loss of function mutation of NF2 occurs in >30% of mesothelioma and in all of a group of benign diseases known as NF2 diseases; similarly, mutation of Hippo components LATS1/2 tumor suppressors also occurs in a variety of cancers. Although genetic mutation of YAP is rare, its overexpression and post-translational activation is frequently observed in cancer. Importantly, malignant mutations of these genes usually drive metastasis, protect cancer cells from apoptosis, and make them more resistant to common cancer therapies^138,139. Therefore, the finding that these same mutations sensitize cancer cells to ferroptosis⁴¹ not only makes these mutations potential biomarkers for ferroptosis-inducing therapies, but also unveils an unusual “Achilles’ heel” of these malignant cells and suggests unique therapeutic opportunity of ferroptosis induction. This opportunity is particularly attractive for gastric cancer and mesothelioma, both of which currently lack effective treatment^140,141.

Potential ferroptosis-inducing cancer therapies

The development of ferroptosis induction-based cancer therapies is being actively pursued. While several untargeted nanoparticle-based strategies to deliver iron, peroxides and other toxic cargoes to kill tumor cells have been tested, the presence of multiple enzymes that control ferroptosis enables the development of targeted approaches^142,143. Perhaps the most obvious target is GPX4, as it is expressed in most cancer cell lines and is important for their survival²². Yet, GPX4 lacks a classical small molecule binding pocket, and the available GPX4 inhibitors covalently modify the selenocysteine residue of GPX4 as well as other selenoproteins¹⁴⁴, raising the issue of specificity. These inhibitors are also highly reactive and thus unstable, but this may be overcome by developing masked prodrugs which can be metabolically converted into their active forms intracellularly^145,146. Nonetheless, the main caveat remains that GPX4 is essential for various peripheral tissues, such as kidney tubular cells and certain neuronal subpopulations in mice²⁹, thus targeting GPX4 will probably cause substantial side effects.

Unlike targeting GPX4, approaches to limit cellular cyst(e)ine availability by inhibiting system x_c^– emerge to be highly promising given the fact that Slc7a11 knockout in mice does not cause major pathologies¹⁴⁷, and that the expression of SLC3A2/SLC7A11 negatively correlates with clinical outcome of melanoma and glioma patients^131,148. Indeed, studies to restrain tumor growth and tumor metastasis in various tumor entities by inhibiting system x_c^– either pharmacologically^149,150 or genetically^151–157 in mouse models have provided highly promising results. The higher vulnerability of various tumor tissues to system x_c⁻ inhibition than that of normal tissues is likely due to the more active metabolism and other alterations in these tumor cells, making them subjective to sustained oxidative stress and thus more addictive to system x_c⁻ function. Obviously, in order to apply system x_c⁻ inhibition-based therapies, a careful stratification of patient tumor tissues is required to examine system x_c⁻ expression (e.g., SLC7A11 overexpression may indicate cancer cell addiction to cystine)⁸² and other biomarkers that determine the sensitivity of tumors to system x_c⁻ inhibition.

Similar to SLC7A11, knockout of FSP1 does not cause embryonic lethality or apparent pathologies¹⁵⁸, suggesting a broad therapeutic window for targeting FSP1. Moreover, FSP1 is abundantly expressed in a large number of cancer cell lines and is the highest ranked gene correlating with the resistance to GPX4 inhibitors in a panel of 860 cancer cell lines⁸². Cancer cells deprived of GPX4 can be efficiently killed by FSP-specific inhibitor iFSP1, while in GPX4-proficient cancer cells iFSP1 synergizes with RSL3⁸². Therefore, FSP1 inhibitors may find their way into the clinics, especially in therapy-resistant or dedifferentiating cancer entities.

Ferroptosis induction-based treatment might also be combined with other therapeutic approaches, with immune checkpoint blockade and radiotherapy as potential options. Combining immune checkpoint blockade with other therapies has been actively pursued. One rationale is that these other therapies may induce immunogenic cell death in tumor tissues to enhance the potency of immune checkpoint blockade. Intriguingly, Zou and colleagues discovered that anti-PDL1 therapy can potentiate ferroptosis-inducing therapy¹³¹. They found that anti-PDL1 antibodies stimulated CD8⁺ T cells to secrete IFNg, and that IFNg subsequently triggered downregulation of both subunits of system x_c⁻ in tumor cells, thus sensitizing cancer cells to ferroptosis. Therefore, immunotherapy in combination with ferroptosis induction represents a promising treatment in that the two therapeutic modalities mutually potentiate each other, leading to synergistic anticancer effect. As of radiation, recent evidence indicates that it can induce ferroptosis on its own, and can synergize with ferroptosis inducers and immunotherapy, providing potentially effective therapeutic combination regimens^43,159,160. This sensitizing effect was observed at the level of cell death, lipid peroxidation, ferroptosis-linked gene expression changes, and lipidomic changes, as well as in cell-line and patient-derived xenograft cancer models and freshly isolated patient glioma slice cultures. Moreover, cytoplasmic but not nuclear radiation was found to synergize with ferroptosis inducers, suggesting the involvement of lipid-peroxidizing activity and glutathione-depleting activity of radiation in ferroptosis, but not its canonical DNA-damaging effects. On the other hand, the ferroptosis-inducing effects of radiation and immunotherapy have been shown to be synergistic, driven by radiation-induced ATM activation and immunotherapy-driven SLC7A11 downregulation¹⁶⁰. Finally, radiation has been shown to upregulate ACSL4 in cancer⁴³. Therefore, the ferroptosis effect of radiation is likely mediated by multiple mechanisms and may be context-dependent. Notably, ferroptosis is implicated as a contributor to some of the adverse events of radiation, such as lung fibrosis and killing of granulocyte-macrophage hematopoietic progenitor cells^161,162.

Ferroptosis in IRI.

Ischemia followed by reperfusion can induce massive cell death and inflammatory response in the affected organs, resulting in devastating diseases including brain stroke, ischemic heart diseases, and injuries of liver and kidney. Remarkably, ischemic heart disease remains to be the disease that causes highest mortality worldwide¹⁶³. Strong evidence, as detailed below, indicates that ferroptosis is a major contributor to IRI-associated cell death, due to, at least partially, oxidative stress induced by ischemia. These findings suggest that ferroptosis inhibition is a potential therapeutic approach for the treatment of IRI diseases.

Brain and heart

Multiple studies have established a role of ferroptosis in neurotoxicity and brain injuries, suggesting the therapeutic potential of ferroptosis inhibition. An ex vivo experiment using rat hippocampal slice culture showed that glutamate-induced neuronal excitotoxic death can be blocked by the ferroptosis inhibitor ferrostatin-1¹. As glutamate neurotoxicity is involved in stroke and various neurodegenerative diseases¹⁶⁴, and high concentration of extracellular glutamate can induce ferroptosis through inhibiting system x_c⁻ function¹, probably ferroptosis contributes to the pathogenesis of these brain diseases. Consistently, genetic studies confirmed that conditional Gpx4 deletion can cause symptoms mimicking neurodegeneration^19,165–167. Further, iron chelators and lipophilic RTAs have been tested for stroke and neurodegeneration in diverse experimental systems^{75,166,168,169}. Although the adverse effect of systemic use of iron chelators remains a serious concern, these experiments have nevertheless validated the significant contribution of ferroptosis in these diseases.

The role of ferroptosis in ischemic heart disease has also been extensively investigated. In an ex vivo system mimicking IRI of mouse heart, it has been shown that iron chelators and glutaminolysis inhibitors significantly mitigated cardiomyocyte cell death and damage of heart tissue and function, suggesting the potential therapeutic value of ferroptosis-targeting for the treatment of ischemic heart disease⁶². More recently, an in vivo mouse model study further confirmed this notion¹⁷⁰.

To design feasible ferroptosis-targeted therapies to treat the fatal diseases of stroke and ischemic heart injury, many important factors need to be considered. For example, both diseases can be extremely acute and only allow a short time window for intervention; therefore, can ferroptosis-targeted therapies be applied timely thereby exerting their effects rapidly enough? Moreover, since there are multiple ferroptosis surveillance pathways, and ischemia/reperfusion may induce ferroptosis in different organs via selectively disrupting one of these surveillance pathways, can we develop specific ferroptosis inhibitors that can effectively treat a certain disease while causing minimal side effect to other organs?

IRI associated with organ transplantation

Besides the brain, perhaps the kidney is the most ferroptosis-sensitive organ identified in adult mammals; survival of the proximal kidney tubular cells depends on functional GPX4⁴⁹, and they are at high risk in response to an IRI scenario as occurring during kidney transplantation. Consistently, ferroptosis inhibitors have been shown to mitigate kidney tubular cell death and acute renal failure in models of warm IRI, in the genetic model of inducible whole body Gpx4 deletion, and in folic acid-induced acute kidney injury (AKI)^49,171,172. These studies suggest that ferroptosis causes activation of the innate immune system via necroinflammation associated with specific molecular alterations^172–174. Another frequently transplanted organ is the liver. While mice with hepatocyte-specific ablation of Gpx4 die neonatally, high dietary vitamin E can compensate for the lack of GPX4¹⁷⁵. Moreover, liproxstatin-1 protects the liver parenchyma from IRI⁴⁹. Lastly, ferroptosis has also been implicated in heart transplantation¹⁷⁶. Collectively, ferroptosis inhibitors might be effective drugs to alleviate tissue detriment inflicted by oxygen deprivation/reoxygenation in different organs.

Potential therapeutic targets for IRI

There are several potential points of therapeutic intervention in the ferroptosis network, although each node of intervention carries its own risk/reward profile. Given that ferroptosis is driven by phospholipid peroxidation, one strategy is to introduce agents that prevent the peroxidation process, for example, by preventing the propagation of lipid peroxyl radicals through the administration of lipophilic RTAs, such as liproxstatin⁵¹. These agents are highly effective in cell models of ferroptosis and can also be effective in some in vivo contexts. Their pharmacological properties need to be further improved before human trials can be considered. A related approach is to administer PUFAs that are chemically resistant to peroxidation, such as by incorporation of deuterium at the bis-allylic carbon that is normally susceptible to peroxidation^37,177. Such deuterated PUFAs have been shown to be effective inhibitors of ferroptosis and a variety of chronic ferroptosis-linked degenerative diseases, although with lower potency than ferrostatin and liproxstatin^37,102. Similarly, monounsaturated fatty acids have been shown to suppress ferroptosis and could be exploited as a potential therapy^36,37.

A second strategy for blocking lipid peroxidation is to either inhibit enzymes that drive lipid peroxidation or to deplete the labile iron pool, e.g., through using iron chelators. Iron chelators have been explored in numerous clinical indications but have selectivity and safety issues to content with. On the other hand, targeting enzymes responsible for generating PUFA-PLs and PLOOHs, such as ACSL4, LPCAT3, LOXs, and POR, might be feasible. As defined molecular targets, they are preferable from a target-centric drug discovery standpoint. However, they also have other functions in various tissues, complicating the potential therapeutic outcomes.

Conclusions and perspectives

Ferroptosis is a concept that integrates into a cohesive and satisfying whole previously disparate aspects involving the metabolism of iron, selenium, amino acids, lipids, and redox chemistry (Fig. 5). This increasingly vibrant tapestry intersects a broad array of biological processes, including both normal physiology and diverse pathophysiologies. From our current vantage point surveying the landscape that is encompassed by the field of ferroptosis, we can share some concerns, challenges, and a preview of potential discoveries to be made in the future.

Figure 5. Relationship of ferroptosis to other biological processes and diseases.

The figure depicts streams of related fields that feed into the interdisciplinary field of ferroptosis, as well as some diseases for which ferroptosis has been implicated. Redox biology, selenium, iron and lipid and amino acid metabolism (streams) are related biological processes and fields that control aspects of ferroptosis. The whirlpools represent diseases that have been linked to ferroptosis: metastatic cancers and drug-resistant cancers have sensitivity to ferroptosis inducers, tumor suppression occurs in some cases through inducing ferroptosis, and neurodegeneration, ischemic organ injury, cardiac injury and organ transplantation involve activation of ferroptosis as part of degenerative processes.

First, we share some concerns. While there is enthusiasm for the increasing explanatory power of the concept of ferroptosis and its interdigitation with diverse aspects of biology, care must be taken to rigorously test the role of ferroptosis in biological processes. Suppression of cell death by a single inhibitor of ferroptosis cannot be taken as sufficient evidence that ferroptosis is involved in a process of interest. The current operational definition of ferroptosis¹⁷⁸ is that of a cell death process suppressed by both iron depletion and lipophilic RTAs, such as ferrostatin-1, liproxstatin-1, vitamin E, or CoQ₁₀. Both requirements are critical, as there are iron-dependent lethal mechanisms distinct from ferroptosis¹⁷⁹ that may involve lysosomal toxicity, for example, and oxidative stress mechanisms independent of iron¹. We suggest adding another requirement: direct detection of lipid peroxidation (using LC-MS/MS, fluorescent dyes, or antibodies such as 1F83). The mobilization and upregulation of TfR1 is another potential marker recently reported^180,181. Although its generality remains to be established, this new marker allows for the differentiation of contexts involving oxidative stress versus ferroptosis. Additional markers of ferroptosis would be highly valuable to the field. Another significant concern is various improper experimental approaches used in ferroptosis research, an inevitable problem in any newly emerging area. To address this issue, Box-2 and Box-3 present a practical guideline for ferroptosis investigation, in which important considerations for in vivo and in vitro study of ferroptosis are described. In particular, we list certain inappropriate tools and approaches that have frequently appeared in the literature but should be avoided in future research.

Box 2 – Potential caveats when using genetically engineered model systems to study ferroptosis.

Sensitivity to ferroptosis is inevitably linked to intra- and extra-cellular redox conditions. For instance, SLC7A11 is upregulated in many cell lines and promptly induced in ex vivo-cultured cells/organs due to the virtual absence of the reduced form of cysteine in culture media. While the knockout of Slc7a11 is well-tolerated in mice because cysteine is present in vivo in its reduced form that is taken up into cells by neutral amino acids transporters¹⁴⁷, almost all cell lines with Slc7a11-knockout invariably die when cultured in vitro as all cysteine is present in its oxidized form in culture medium. These stark discrepancies in redox conditions between cell culture-based and in vivo studies call for a careful interpretation of findings solely based on in vitro studies. Equally important, when using system x_c⁻ inhibitors it needs to be assessed whether SLC7A11 is expressed at all in the cell line to be studied. Alarmingly, many currently available antibodies against SLC7A11 are unspecific and unsuitable for determining SLC7A11 levels.

When intervening with GPX4 activity, one needs to consider that GPX4 is promiscuous not only using glutathione as substrate, but also other thiol-containing compounds and even protein thiols²⁶. This is of particular significance when using ferroptosis-inducing agents that act at the different steps of the cyst(e)ine-glutathione-GPX4 axis. Moreover, since cell density has a profound impact on the cells’ sensitivity toward ferroptosis^19,41, it needs to be carefully controlled. Notably, since GPX4 requires selenium for its biosynthesis and sera contain varying amounts of selenium (and vitamin E), different cell culture sera may have a profound impact on the results. This is also of vital importance when using Gpx2^−/− mice since chows can vary in their vitamin E contents (both tocopherols and tocotrienols), which have major impact on the phenotypes of knockout mice^175,190,191. Therefore, this should also be considered in future studies when using pathological models or tumor xenografts. Consistently, supplementation with selenium and vitamin E has been shown to be cancer-promoting in some contexts¹⁹².

Box 3 – Experimental consideration when using pharmacological tools to probe ferroptosis.

Small molecule tools are valuable for probing ferroptosis. However, several issues involving specificity, biological relevance, pharmacokinetics and pharmacodynamics must be considered when using such tools. First, when using pharmacological inhibitors of a certain enzyme to assess whether the enzyme regulates ferroptosis, one needs to determine whether the purported inhibitor can act as lipophilic RTAs and thus inhibit ferroptosis independent of the enzyme under study, as occurred in some LOX-ferroptosis studies^51,52,193. Second, it is important to verify in the experimental system whether a given small molecule probe indeed affects the intended target or pathway. For example, in using RSL3 to inhibit GPX4, or erastin to inhibit system x_c⁻, it is important to test whether those targets are inhibited by each probe using a biochemical assay or a pharmacodynamic marker of target inhibition, such as mRNA expression changes. Moreover, when using small molecule in animal studies, it is important to first determine the appropriate formulation and route of administration to achieve and acceptable pharmacokinetic profile and pharmacodynamic response of the target tissue. Many small molecules that are potent and selective probes in cellular assays have no utility in animal models due to poor solubility, stability and/or pharmacokinetics. A few common approaches that are inappropriate and thus should be avoided are listed below:

• RSL3 has low solubility and difficult to measure pharmacokinetics in mice, not suitable suited for in vivo animal studies, except in some cases by direct injection into tissues or tumors. Erastin has low solubility and low metabolic stability, unsuitable for use in animal studies (imidazole ketone erastin (IKE) has good solubility, high stability and favorable pharmacokinetics and pharmacodynamics in mice, and has been validated for in vivo studies). Similarly, ferrostatin-1 is not suitable for in vivo studies, whereas liproxstatin-1 and vitamin E can be used in vivo as ferroptosis inhibitors. In vivo data generated by using iron chelators should be treated with caution, as depleting iron may lead to a plethora of consequences in addition to ferroptosis inhibition, and it is unclear whether a specific outcome is caused by iron chelation or an off-target activity of the compounds. Inducing or potentiating ferroptosis in cell culture by adding iron (ferric or ferrous) into the culture medium is not appropriate for the study of ferroptosis mechanisms, as free iron ions catalyze Fenton chain reaction in the presence of oxygen and trace amount of lipid peroxides (generated by accidental cell death or ferroptosis triggered by other inducers), thus amplifying extracellular lipid peroxidation and killing cells rapidly.

Second, we note a major challenge. Despite significant progress in understanding the mechanisms regulating ferroptosis, we still do not know how cells ultimately die. Uncontrolled peroxidation of PUFA-PLs is the most downstream step identified; it may be that peroxidized phospholipids cause membrane damage or even pore formation, compromising membrane integrity. However, a recent report found that phospholipids containing two PUFA tails are particularly effective drivers of ferroptosis⁸⁹, suggesting that crosslinking may be an aspect of the membrane damage upon ferroptosis. In this scenario, it may be that limiting fluidity of membrane components causes failure of some vital functions associated with membranes, resulting in cell death (analogous to vascular stiffening seen in vivo). However, other possibilities exist, such as that oxidized PUFA-PLs decompose into reactive electrophiles that then damage other macromolecules. In this model, reactive electrophiles serve the function in ferroptosis that caspases serve in apoptosis—to inactivate key structural and functional proteins within cells. We hope that over the next several years, the mechanism of ferroptosis execution will be unambiguously elucidated.

Third and finally, we can predict some potential future discoveries that lie ahead in the field of ferroptosis. Identifying precise biomarkers for in vivo ferroptosis, as cleaved caspase-3 for apoptosis in vivo and in tissue sections, will be fundamentally important for determining the physiological function and therapeutic role of this cell death modality. Identifying the enigmatic natural triggers for ferroptosis also remains a pressing challenge, but one that may see breakthroughs in the coming years. The following molecules are potentially, or proximate to, the natural ferroptosis triggers—glutamate, p53, iron, radiation, and PUFA-PLs. Glutamate is an abundant neurotransmitter, and excess extracellular glutamate is sufficient to inhibit system x_c⁻, resulting in ferroptosis¹. Nature may exploit this effect of extracellular glutamate to induce ferroptosis in developmental or other physiological contexts, such as the elimination of unnecessary neurons in the sculpting of neural circuits. The tumor suppressor p53 induces ferroptosis likely to suppress tumor development; induction of p53 in the presence of basal lipid peroxidative damage may be a natural means of eliminating certain stressed cells through ferroptosis. High levels of iron alone can trigger ferroptosis in diverse model systems, suggesting that in some contexts, releasing ferritin-stored iron or increasing iron import might be a means of eliminating cells through ferroptosis. Low to moderate doses of ionizing radiation have been shown to induce ferroptosis. Thus, contexts in which radiation is abundant, or radioprotective cellular mechanisms are compromised might be circumstances in which ferroptosis is naturally induced. For example, there is speculation that there might once have been life on Mars, which either has been lost or has moved underground due to the harsh radiation present on the planet surface. Thus, high radiation environments might make ferroptosis a default program that needs to be continuously counteracted for life to survive and evolve, by establishing a network of anti-ferroptotic mechanisms; and downregulating these protective mechanisms would then serve as a natural trigger for driving ferroptosis. Finally, PUFA uptake and PUFA-PL synthesis may be sufficient in some contexts to induce ferroptosis; in such cases, a cell could be induced to die by ferroptosis by release of PUFAs in the extracellular environment, or by driving expression of enzymes such as LPCAT3 and ACSL4 that stimulate PUFA-PL synthesis. Some of these mechanisms may be involved in development, tumor suppression and clearance of infected cells by the immune system, and regulation of tissue size.

In summary, there is a wealth of foreseeable opportunity to elucidate both the execution mechanisms of ferroptosis and the contexts in which this form of cell death is naturally harnessed. Such studies will illuminate how Nature chooses to co-opt ferroptosis to an array of purposes beyond disease and therapy.

Glossary:

Alkoxyl radicals

a reactive species consisting of a hydrocarbon chain attached to an oxygen with an unpaired electron

Anaplerosis

replenishment of TCA cycle intermediates

Bis-allylic carbon

a carbon that is in between and adjacent to two carbon-carbon double bonds

Conjugated diene

two carbon-carbon double bonds separated by a carbon-carbon single bond, such that the pi systems of the two double bonds form a continuous delocalized pi system

Dihydrobiopterin (BH₂)

a natural biopterin metabolite in the partially oxidized form

Immune checkpoint blockade

the process of activating immune cells by inhibiting a signaling pathway that normally suppresses their activity

Ischemia/reperfusion injury

the organ injury resulting from limited blood flow and oxygen deprivation followed by reintroduction of blood flow and oxygenation

Mevalonate pathway

a lipid biosynthesis pathway that produces terpenes-derived compounds such as cholesterol and coenzyme Q₁₀

Selenoproteins

proteins into which the element selenium is incorporated, mainly via the incorporation of selenocysteine

Tetrahydrobiopterin (BH₄)

a natural biopterin metabolite in the fully reduced form

Transferrin

a protein that transports iron into cells through receptor-mediated endocytosis