Targeting ferroptosis as a vulnerability in cancer

Abstract

Ferroptosis is an iron-dependent form of regulated cell death that is triggered by the toxic build-up of lipid peroxides on cellular membranes. In recent years, ferroptosis has garnered enormous interest in cancer research communities, partly because it is a unique cell death modality that is mechanistically and morphologically different from other forms of cell death such as apoptosis and therefore, holds great potential for cancer therapy. In this Review, we summarize the current understanding of ferroptosis-inducing and ferroptosis defence mechanisms, dissect the roles and mechanisms of ferroptosis in tumour suppression and tumour immunity, conceptualize the diverse vulnerabilities of cancer cells to ferroptosis, and explore therapeutic strategies for targeting ferroptosis in cancer.

summary

In recent years, research in the field of ferroptosis in cancer has risen steeply and this is in part owing to its potential to be targeted. In this Review, Lei et al. provide an up-to-date synthesis of the roles and mechanisms of ferroptosis in tumour growth and progression including its function in tumour immunity, highlighting it as a vulnerability that can be exploited for cancer therapy.

Introduction

Ferroptosis, a term coined by Stockwell and colleagues in 2012¹, refers to an iron-dependent form of regulated cell death driven by an overload of lipid peroxides on cellular membranes. It is morphologically and mechanistically distinct from apoptosis and other types of regulated cell death; for example, morphologically, cells undergoing ferroptosis do not exhibit typical apoptotic features (such as chromatin condensation and apoptotic body formation), but are characterized by shrunken mitochondria and reduced numbers of mitochondrial cristae^1,2. The lethal accumulation of lipid peroxides is a cardinal feature of ferroptosis³ and involves an antagonism between ferroptosis execution and ferroptosis defence systems in cells; ferroptosis occurs when ferroptosis-promoting cellular activities significantly override the antioxidant-buffering capabilities provided by ferroptosis defence systems ^4–10 (Fig. 1a). This mechanistic feature is markedly different from several other forms of regulated cell death that centre on cell death executioner proteins (such as caspase-mediated apoptosis, mixed lineage kinase domain-like protein [MLKL]-mediated necroptosis, and gasdermin D-mediated pyroptosis)¹¹. Finally, ferroptotic cells exhibit distinctive oxidized phospholipid (PL) profiles which differ from cells undergoing other forms of cell death^12,13.

Figure 1. Ferroptosis-driving and -defence mechanisms.

a. Ferroptosis reflects an antagonism between prerequisites for ferroptosis and ferroptosis-defence systems. The prerequisites for ferroptosis consist of polyunsaturated fatty acid-containing phospholipid (PUFA-PL) synthesis and peroxidation, iron metabolism and mitochondrial metabolism. Ferroptosis-defence systems mainly include the glutathione peroxidase 4 (GPX4)–reduced glutathione (GSH) system, the ferroptosis suppressor protein-1 (FSP1)–ubiquinol (CoQH₂) system, the dihydroorotate dehydrogenase (DHODH)–CoQH₂ system, and the GTP cyclohydroxylase-1 (GCH1)– tetrahydrobiopterin (BH₄) system. When ferroptosis-promoting cellular activities significantly exceed the detoxification capabilities provided by ferroptosis defence systems, a lethal accumulation of lipid peroxides on cellular membranes lead to subsequent membrane rupture and ferroptotic cell death.

b. Acyl-coenzyme A synthetase long chain family member 4 (ACSL4) and lysophosphatidylcholine acyltransferase 3 (LPCAT3) mediate the synthesis of PUFA-PLs, which are susceptible to peroxidation through both nonenzymatic and enzymatic mechanisms. Iron initiates the nonenzymatic Fenton reaction and acts as an essential cofactor for arachidonate lipoxygenases (ALOXs) and cytochrome P450 oxidoreductase (POR), which promote lipid peroxidation, and mitochondrial metabolism promotes the generation of reactive oxygen species (ROS), ATP, and/or PUFA-PLs. Excessive accumulation of lipid peroxides on cellular membranes can trigger ferroptosis. Cells have evolved at least 4 defence systems with different subcellular localizations to detoxify lipid peroxides and thus protect cells against ferroptosis, wherein cytosolic GPX4 (GPX4^cyto) cooperates with FSP1 on the plasma membrane (and other non-mitochondrial membranes) and mitochondrial GPX4 (GPX4^mito) with DHODH in the mitochondria to neutralize lipid peroxides. The subcellular compartment in which the GCH1–BH₄ system operates remains to be defined. CoQ, coenzyme Q (also known as ubiquinone).

Ferroptosis as a unique cell death mechanism has sparked great interest in the cancer research community, as targeting ferroptosis might provide new therapeutic opportunities in treating cancers that are refractory to conventional therapies. In recent years, substantial progress has been achieved in understanding the role of ferroptosis in tumour biology and cancer therapy. On one hand, multiple cancer-associated signalling pathways have been shown to govern ferroptosis in cancer cells¹⁴. The engagement of ferroptosis in the activities of several tumour suppressors, such as p53 and BRCA1-associated protein 1 (BAP1), establishes ferroptosis as a natural barrier to cancer development^15,16, whereas oncogene-mediated or oncogenic signalling-mediated ferroptosis evasion contributes to tumour initiation, progression, metastasis and therapeutic resistance^17–19. On the other hand, the distinctive metabolism of cancer cells, their high load of reactive oxygen species (ROS), and their specific mutations render some of them intrinsically susceptible to ferroptosis, thereby exposing vulnerabilities that could be therapeutically targetable in certain cancer types^20–24. Furthermore, some cancer cells appear to be particularly dependent on ferroptosis defence systems to survive under metabolic and oxidative stress conditions; consequently, disruption of those defences would be fatal to such cancer cells while sparing normal cells⁹. These recent data suggest that ferroptosis represents a targetable vulnerability of cancer in certain contexts. Ferroptosis has also been recognized as a critical cell death response triggered by a variety of cancer therapies, including radiotherapy (RT), immunotherapy, chemotherapy, and targeted therapies^25–28. Thus, ferroptosis inducers (FINs) hold great potential in cancer therapy (Box 1), especially in combination with conventional therapies^25,29,30.

Box 1. Ferroptosis inducers.

Ferroptosis inducers (FINs) are compounds or treatments that induce ferroptosis in cells, and can be categorized into at least four classes based on their mechanisms of action¹⁷⁰. Class I FINs act by inhibiting solute carrier family 7 member 11 (SLC7A11)-mediated cystine uptake and restricting intracellular cysteine and glutathione levels^82,170. Erastin is the most widely used class I FIN in cell culture studies, although its application in vivo is limited by its poor metabolic stability and water solubility¹⁷⁰. Its analogue imidazole ketone erastin (IKE) has improved potency, solubility, and metabolic stability¹⁷¹, and has been shown to induce ferroptosis and suppress tumour growth in vivo ²⁹. Sulfasalazine, a US Food and Drug Administration (FDA)-approved anti-inflammatory agent with class I FIN activity^172,173, can inhibit tumour growth in diverse preclinical models^97,173. Although sulfasalazine exhibits relatively poor metabolic stability and potency¹⁷⁰, its activity in patients with glioblastoma has been demonstrated¹⁷⁴ and its combination with radiotherapy is being tested in a clinical trial for glioblastoma (NCT04205357¹⁶¹). Sorafenib, another clinically approved agent is a multi-kinase inhibitor with activity against SLC7A11¹⁴⁴; although it has been shown to induce ferroptosis, it also induces other nonferroptotic cellular effects^144,170, and its classification as a FIN was challenged by a recent study¹⁷⁵. In addition, cyst(e)inase degrades extracellular cystine and cysteine and has been shown to act as a potent and tolerable FIN to suppress tumour growth in vivo, when used either as a treatment alone or in combination with immune checkpoint inhibitors^28,176.

Class II FINs operate by blocking the enzymatic activity of glutathione peroxidase 4 (GPX4), and include RSL3, ML162, and ML210. RSL3 and ML162 directly inhibit the catalytic activity of GPX4 by covalently binding to the selenocysteine residue of GPX4 through their electrophile chloroacetamide moiety^4,165; however, their application in vivo is limited by their low solubility and poor pharmacokinetics^165,170. ML210 is a nitroisoxazole-containing compound that is converted to α-nitroketoxime (known as JKE-1674) in cells and subsequently forms a nitrile-oxide electrophile, which covalently inhibits GPX4 with remarkable selectivity¹⁶⁵. Notably, JKE-1674 exhibits improved pharmacokinetic properties¹⁶⁵, although its effect on tumour growth remains to be investigated. Moreover, altretamine (an FDA-approved anticancer agent) and withaferin A (a natural anticancer agent suitable for in vivo treatment) have been shown to inhibit GPX4 activity or decrease GPX4 levels, providing an alternative for targeting GPX4 in vivo^166,167.

Finally, class III FINs, such as FIN56³⁶, act by depleting both GPX4 protein and ubiquinone. Class IV FINs, such as FINO₂¹⁷⁷, act by oxidizing iron and indirectly inactivating GPX4. However, these FINs have not been assessed in vivo.

With the exponential growth in ferroptosis research in the past 4 years, iterative insights into how to target ferroptosis in cancer are critically needed. In this Review, we summarize the current understanding of the regulatory networks of ferroptosis, including the prerequisites for ferroptosis and ferroptosis defence systems, thoroughly analyse the mechanistic bases of ferroptosis in tumour biology and synthesize a conceptual framework for how to target ferroptosis as a vulnerability in cancer therapy. Lastly, we highlight several key questions and challenges for future studies.

Ferroptosis prerequisites

The crux of ferroptosis execution is iron-catalysed peroxidation of polyunsaturated fatty acid (PUFA)-containing PLs (PUFA-PLs), which, when exceeding the buffering capability of ferroptosis defence systems (see the next section), can lead to lethal accumulation of lipid peroxides on cellular membranes and subsequent membrane rupture, resulting in ferroptotic cell death. As outlined in this section, PUFA-PL synthesis and peroxidation, iron metabolism, and mitochondrial metabolism constitute the main prerequisites driving ferroptosis^3,31 (Fig. 1a).

PUFA-PL synthesis and peroxidation.

Acyl-coenzyme A (CoA) synthetase long chain family member 4 (ACSL4) and lysophosphatidylcholine acyltransferase 3 (LPCAT3) are critical mediators of PUFA-PL synthesis^13,32,33 (Fig. 1b). ACSL4 catalyses the ligation of free PUFAs, such as arachidonic acids and adrenic acids, with CoA to generate PUFA-CoAs (such as arachidonic acid-CoA or adrenic acid-CoA)^32,33, which are subsequently re-esterified and incorporated into PLs by LPCAT3 to form PUFA-PLs (such as arachidonic acid-phosphatidylethanolamine [PE] or adrenic acid-PE)^13,32. Acetyl-CoA carboxylase (ACC)-catalysed carboxylation of acetyl-CoA generates malonyl-CoA, which is required for the synthesis of some PUFAs (likely at the elongation steps) and therefore for ferroptosis^32,34–36. Inactivation of ACSL4, LPCAT3, or ACC blocks or attenuates ferroptosis^13,33,37. Interestingly, p53-mediated ferroptosis, in some contexts, seems to be independent of ACSL4³⁸, although how ferroptosis occurs in the absence of ACSL4-mediated PUFA-PL synthesis remains unclear. Recent findings also revealed that PUFA-containing ether PLs (PUFA-ePLs) synthesized in peroxisomes act as additional substrates for lipid peroxidation, favouring ferroptosis onset^23,39,40.

PUFA-PLs are particularly susceptible to peroxidation due to the presence of bis-allylic moieties in PUFAs⁴¹. The peroxidation of PUFA-PLs is primarily catalysed by nonenzymatic autoxidation driven by the Fenton reaction with iron as a catalyst^41–43 (Fig. 1b). Cytochrome P450 oxidoreductase (POR)- or arachidonate lipoxygenase (ALOX)-mediated enzymatic reactions have also been shown to promote lipid peroxidation^44–48; although the ability of POR to promote lipid peroxidation appears to be indirect via the generation of hydrogen peroxide (H₂O₂)⁴⁶ and the role of ALOXs in ferroptosis has been challenged by other studies, including the lack of rescue of glutathione peroxidase 4 (Gpx4; part of a ferroptosis defence mechanism) knockout phenotypes by Alox15 deletion in mice⁴⁹. Monounsaturated fatty acids (MUFAs), such as oleic acid and palmitoleic acid, are not readily peroxidised owing to their lack of bis-allylic moieties⁵⁰. In contrast to PUFAs, MUFAs restrain lipid peroxidation and ferroptosis by displacing PUFAs from PLs in cellular membranes; inactivation of enzymes involved in MUFA-PL synthesis, such as stearoyl CoA desaturase 1 (SCD1) and ACSL3, sensitizes cancer cells to ferroptosis^50–52.

Iron metabolism.

Iron governs ferroptosis not only by initiating the nonenzymatic Fenton reaction for direct peroxidation of PUFA-PLs^41,43, but also by acting as an essential cofactor for enzymes that participate in lipid peroxidation (such as ALOX and POR)^44,45. Cells normally maintain a relatively stable pool of labile iron through orchestrated regulation of iron uptake, utilization, storage, and export⁵³. Imbalanced regulation of these iron metabolism processes can promote or suppress ferroptosis, depending on whether the level of the intracellular labile iron pool is increased or decreased, respectively. For example, intracellular iron is stored mostly as inert iron in ferritin, whereas autophagic degradation of ferritin (ferritinophagy) releases iron stored in ferritin into the labile iron pool; consequently, blockade of nuclear receptor coactivator 4 (NCOA4)-mediated ferritinophagy decreases the level of the labile iron pool and suppresses ferroptosis^54,55. Conversely, enhancement of ferritinophagy by inhibition of cytosolic glutamate oxaloacetate transaminase 1 (GOT1) increases the labile iron pool and promotes ferroptosis (although exactly how GOT1 inhibition promotes ferritinophagy-mediated labile iron release remains to be further studied)⁵⁶. We refer readers to a recent excellent review for detailed discussion of iron metabolism in ferroptosis⁵³.

Mitochondrial metabolism.

Several metabolic processes in mitochondria have important roles in triggering ferroptosis³¹ (Fig. 1b). First, the generation of mitochondrial ROS is critical for lipid peroxidation and ferroptosis onset. Mitochondria are a major source of cellular ROS, in which electron leakage from electron transport chain complexes I and III generates superoxides, followed by their conversion to H₂O₂ by superoxide dismutase⁵⁷. H₂O₂ can then react with labile iron via the Fenton reaction to generate hydroxyl radicals, which subsequently drive PUFA-PL peroxidation^57,58. Moreover, electron transport and proton pumping in mitochondria are important for ATP production^59,60, which also promotes ferroptosis^35,37. Specifically, under ATP-depleted conditions, AMP-activated protein kinase (AMPK) phosphorylates and inactivates ACC, thereby suppressing PUFA-PL synthesis and blocking ferroptosis; in contrast, with sufficient energy (i.e., ATP), AMPK cannot be activated efficiently and ACC is activated, thus promoting PUFA-PL synthesis and ferroptosis^35,37. Finally, the role of mitochondria in biosynthetic pathways in cellular metabolism also contributes to ferroptosis. Ferroptosis requires the tricarboxylic acid (TCA) cycle and various anaplerotic reactions that fuel the TCA cycle (such as glutaminolysis) in mitochondria⁵⁹, which drive ferroptosis likely through promoting ROS, ATP, and/or PUFA-PL generation^61–63. Therefore, the multifaceted functions of mitochondria in bioenergetics, biosynthesis, and ROS generation drive mitochondrial lipid peroxidation and ferroptosis³¹.

Ferroptosis defence mechanisms

Ferroptosis defence mechanisms involve cellular antioxidant systems that directly neutralize lipid peroxides. Studies of ferroptosis defence mechanisms have been one of the most exciting and rapidly evolving areas in ferroptosis research in the past 2 years. As discussed below, there are at least 4 such ferroptosis defence systems with distinctive subcellular localizations (Fig. 1).

The GPX4–GSH system.

GPX4 belongs to the GPX protein family^64,65 and is the only GPX member capable of converting PL hydroperoxides to PL alcohols^66,67. Genetic ablation or pharmacological inhibition of GPX4 induces unchecked lipid peroxidation and triggers potent ferroptosis under many in vitro and in vivo conditions^4,49,66. GPX4 consists of 3 isoforms with distinctive subcellular localizations, namely cytosolic, mitochondrial, and nuclear GPX4s. These isoforms are encoded by the same GPX4 gene with different transcription initiation sites, resulting in the addition of a mitochondrial or nuclear localization sequence to the N-terminus of the GPX4 protein^68–71. Until very recently, it was believed that only cytosolic GPX4 had a role in defending against ferroptosis, because cytosolic GPX4, but not the mitochondrial or nuclear isoform, is required for embryonic development^72–75 and because reexpression of cytosolic GPX4, but not mitochondrial or nuclear GPX4, significantly suppresses cell death induced by Gpx4 deletion in mouse embryonic fibroblasts⁷⁶. However, recent studies suggest that both cytosolic and mitochondrial GPX4 are important in defending against ferroptosis in different subcellular compartments (see below)⁹. The potential role of nuclear GPX4 in the regulation of ferroptosis remains to be studied.

Reduced glutathione (GSH), the cofactor used by GPX4, is a thiol-containing tripeptide that is derived from glycine, glutamate, and cysteine, with cysteine being the rate-limiting precursor^77,78. Most cancer cells obtain intracellular cysteine primarily through system x_c⁻-mediated uptake of cystine (an oxidized dimeric form of cysteine), followed by cystine reduction to cysteine in the cytosol^79,80. Solute carrier family 7 member 11 (SLC7A11; also known as xCT) is the transporter subunit in system x_c^{− 81,82}. Removing cystine from culture media or pharmacologically blocking SLC7A11-mediated cystine transport with erastin or other FINs induces potent ferroptosis in many cancer cell lines^1,15,16,82. The SLC7A11–GSH–GPX4 axis is believed to constitute the major cellular system defending against ferroptosis (Fig. 1b). However, upon GPX4 inactivation, some cancer cell lines remain resistant to ferroptosis²², suggesting the existence of additional ferroptosis defence mechanisms.

The FSP1–CoQH₂ system.

While it was initially believed that GPX4 was the only ferroptosis defence system, this paradigm shifted following recent studies revealing that ferroptosis suppressor protein-1 (FSP1; also known as AIFM2) operates independently of GPX4 to defend against ferroptosis^5,6. FSP1 localizes on the plasma membrane (as well as other subcellular compartments) and its plasma membrane localization appears to be both necessary and sufficient for the function of FSP1 in suppressing ferroptosis^5,6. FSP1 functions as an NAD(P)H-dependent oxidoreductase capable of reducing ubiquinone (also known as coenzyme Q or CoQ)^83,84 to ubiquinol (CoQH₂)^5,6 (Fig. 1b). Apart from its well-known function in mitochondrial electron transport, CoQH₂ can also trap lipid peroxyl radicals, thereby suppressing lipid peroxidation and ferroptosis. Therefore, it has been proposed that FSP1 exerts its potent antiferroptosis activity through generating the nonmitochondrial CoQH₂ pool as radical-trapping antioxidants^5,6. It should be noted that CoQ is mainly synthesized in mitochondria⁸³ but has been detected in nonmitochondrial membranes, including the plasma membrane^85–88. The sources of the nonmitochondrial CoQ utilized by FSP1 for ferroptosis defence remain to be established.

The DHODH–CoQH₂ system.

A recent study uncovered a mitochondria-localized defence system mediated by dihydroorotate dehydrogenase (DHODH) that can compensate for GPX4 loss to detoxify mitochondrial lipid peroxidation⁹. DHODH is an enzyme involved in pyrimidine synthesis that can reduce CoQ to CoQH₂ in the inner mitochondrial membrane (Fig. 1b). When GPX4 is acutely inactivated, the flux through DHODH is significantly increased, resulting in enhanced CoQH₂ generation that neutralises lipid peroxidation and defends against ferroptosis in mitochondria⁹. Consequently, inactivation of both mitochondrial GPX4 and DHODH unleashes potent mitochondrial lipid peroxidation and triggers robust ferroptosis. Importantly, while mitochondrial GPX4 and DHODH can compensate for each other to suppress mitochondrial lipid peroxidation, cytosolic GPX4 and FSP1 fail to do so⁹, ostensibly because they are not localized in mitochondria and therefore cannot detoxify lipid peroxides accumulated in the inner mitochondrial membrane, highlighting the importance of compartmentalization in ferroptosis defence.

These recent studies suggest a model whereby ferroptosis defence systems can be divided into 2 major divisions, the GPX4 system and the CoQH₂ system, with each of these further subdivided into arms located in nonmitochondrial and mitochondrial compartments (i.e., cytosolic and mitochondrial GPX4 in the GPX4 system, versus FSP1 and DHODH in the CoQH₂ system). The rationale for this division in ferroptosis defence is likely derived from the significant need to mitigate lipid peroxides generated in mitochondria (see subsection on mitochondrial metabolism) and the double-membrane structure of mitochondria (which prohibits cytosol- or other compartment-localized defence systems from entering into mitochondria). Further research is required to substantiate this compartmentalization model in ferroptosis regulation and to reconcile with some conflicting data from previous studies. For example, cytosolic GPX4 has also been found to significantly localize in the intermembrane space of mitochondria⁷⁴. The abundance of cytosolic GPX4 in mitochondria and its potential role in suppressing lipid peroxidation within mitochondria remain to be clarified by future investigations.

Another open question concerns potential roles of other mitochondrial enzymes that produce CoQH₂, most notably electron transport chain complexes I and II, in ferroptosis regulation. However, unlike DHODH inactivation, inactivation of complex I or II by pharmacological inhibition does not appear to affect GPX4 inactivation-induced ferroptosis or even promotes resistance to cystine starvation-induced ferroptosis⁶². Conceivably, the presumed antiferroptosis function of complex I or II (by producing mitochondrial CoQH₂) might be offset by its proferroptosis functions (through producing ROS or ATP; see subsection on mitochondrial metabolism). The potential role of additional CoQH₂-producing enzymes localized in mitochondria, such as electron transfer flavoprotein dehydrogenase (ETFDH), in ferroptosis regulation awaits further research.

The GCH1–BH₄ system.

Recent studies identified GTP cyclohydroxylase-1 (GCH1) as another critical regulator of ferroptosis^7,8. Tetrahydrobiopterin (BH₄) is a cofactor of aromatic amino acid hydroxylases and other enzymes, and GCH1 mediates the rate-limiting reaction in the BH₄ biosynthesis pathway⁸⁹. BH₄ is another radical-trapping antioxidant capable of trapping lipid peroxyl radicals, and its function in inhibiting ferroptosis appears to be independent of its role as a cofactor⁸. It was proposed that GCH1 suppresses ferroptosis through generating BH₄ as a radical-trapping antioxidant as well as via GCH1-mediated production of CoQH₂ and PLs containing two PUFA tails^7,8 (whereas most PUFA-PLs tend to exhibit an asymmetric structure, with a PUFA tail at the sn-2 position and a saturated fatty acid (SFA) tail at the sn-1 position) (Fig. 1b). The subcellular compartment wherein the GCH1–BH₄ system operates remains to be defined.

Ferroptosis evasion fuels tumours

Accumulating evidence indicates that ferroptosis is a critical tumour suppression mechanism. Several tumour-suppressor and oncogenic signalling pathways have been shown to promote or suppress ferroptosis, respectively (Boxes 2 and 3). Tumours have evolved at least 3 mechanisms to evade ferroptosis and to facilitate tumour development and metastasis, including limiting PUFA-PL synthesis and peroxidation, restricting labile iron availability, and upregulating cellular defence systems against ferroptosis.

Box 2. Ferroptosis regulation by tumour suppressors.

p53, the most frequently mutated tumour suppressor¹⁷⁸, has an important role in regulating ferroptosis^15,179. p53 inhibits solute carrier family 7 member 11 (SLC7A11) expression through binding directly to the SLC7A11 promoter or via interacting with ubiquitin-specific processing protease 7 (USP7) to reduce histone H2B monoubiquitination levels on the SLC7A11 promoter, sensitizing cancer cells to ferroptosis in an arachidonate 12-lipoxygenase (ALOX12)-dependent manner^{15,38,142,180}. p53 can also promote ferroptosis by modulating additional metabolic targets^181–183. Importantly, some p53 mutants cannot induce apoptosis, senescence or cell cycle arrest, yet can promote ferroptosis to mediate tumour suppression in vivo¹⁵. In contrast, other p53 mutants lacking ferroptosis regulatory activity lose their tumour-suppression function^183,184. Note that p53 appears to play a dual role in ferroptosis regulation depending on the cellular context, as it can exert an antiferroptosis effect under certain conditions^185,186.

The tumour suppressor BRCA1-associated protein 1 (BAP1) encodes a nuclear deubiquitinating enzyme (DUB) that reduces histone H2A ubiquitination (H2Aub) on chromatin^187,188. BAP1 is frequently mutated in several sporadic cancers, including uveal melanoma, mesothelioma, and renal cell carcinoma (RCC)¹⁸⁹. BAP1 reduces H2Aub occupancy on the SLC7A11 promoter and represses SLC7A11 expression in a DUB-dependent manner, thereby inhibiting cystine uptake and promoting ferroptosis, which partly mediates the tumour-suppressive effect of BAP1 in xenograft models¹⁶. In addition, the ability of BAP1 to restrain SLC7A11 expression and promote ferroptosis is compromised in cells with cancer-associated BAP1 mutations¹⁶.

Kelch-like ECH associated protein 1 (KEAP1) is a tumour suppressor frequently mutated in non-small-cell lung cancer^190,191. It encodes a substrate adaptor that targets the transcription factor nuclear factor erythroid 2-related factor 2 (NRF2) for proteasomal degradation^192,193. KEAP1 inactivation in cancer leads to NRF2 protein stabilization^192,193. As a master regulator of antioxidant defence, NRF2 promotes the transcription of many genes governing ferroptosis suppression, such as SLC7A11^99,101,194. In addition, ARF is a well-established tumour suppressor that is critical for p53 activation in response to oncogenic stress¹⁹⁵. Interestingly, ARF was recently shown to inhibit NRF2 transcriptional activity independently of KEAP1 or p53¹⁰⁰. In ARF-mutated tumours, NRF2 and its downstream target genes (such as SLC7A11) are activated, enabling cancer cells to evade ferroptosis under oxidative stress¹⁰⁰. Therefore, ferroptosis activation by KEAP1 or ARF due to NRF2 suppression is likely a tumour-suppressive mechanism.

Most studies have focused on ferroptosis regulation mediated by tumour suppressors through SLC7A11. The control of ferroptosis by other tumour suppressors through other mechanisms remains to be explored.

Box 3. Ferroptosis regulation by oncogenes.

In line with the concept that ferroptosis is a natural tumour-suppressor mechanism, several oncogene-driven cancers have evolved mechanisms to escape from ferroptosis.

The PI3K oncogenic signalling pathway is one of the most frequently mutated pathways in human cancers^196,197. It was recently demonstrated that cancer cells carrying PIK3CA activating mutations or PTEN deletion are generally resistant to ferroptosis¹⁷. Oncogenic activation of the PI3K signalling pathway promotes the activation of mTOR complex 1 (mTORC1)¹⁹⁷, which inhibits ferroptosis through at least 3 parallel mechanisms¹⁹⁸: promoting glutathione peroxidase 4 (GPX4) protein synthesis¹²⁰, suppressing autophagy-dependent ferroptosis¹⁹⁹ and boosting monounsaturated fatty acid-containing phospholipid (MUFA-PL) synthesis by increasing stearoyl coenzyme A desaturase 1 (SCD1) expression¹⁷. Thus, aberrant activation of mTORC1 allows cancer cells to evade ferroptosis, which might partly account for the oncogenic activity caused by mutations in the PI3K pathway.

The RAS family of proto-oncogenes are frequently mutated in human cancers^200,201. Notably, RAS-mutant cancer cells are characterized by high levels of intracellular cysteine and glutathione in response to oxidative stress, via mechanisms involving solute carrier family 7 member 11 (SLC7A11) upregulation¹⁰³. Furthermore, pharmacological or genetic inhibition of SLC7A11 severely impairs mutant KRAS-induced tumour growth, suggesting that SLC7A11 is required for mutant KRAS-mediated tumourigenesis^102,103. Therefore, SLC7A11-mediated cystine uptake and glutathione biosynthesis allow mutant RAS-driven tumours to evade ferroptosis under oxidative stress. It is also worth mentioning that KRAS mutation in lung cancer also increases Acyl-coenzyme A synthetase long chain family member 3 (ACSL3) expression to reprogram lipid metabolism, promoting MUFA-PL biosynthesis and ferroptosis resistance, and that these processes might be important for the development of mutant KRAS-driven lung cancer^50,202.

Several noncanonical oncogenes also contribute to the escape of cancer cells from ferroptosis. Neural precursor cell expressed, developmentally down-regulated 4 (NEDD4), a ubiquitin ligase with oncogenic functions²⁰³, can be induced by the ferroptosis inducer erastin¹⁶⁹. Erastin-induced ferroptosis requires voltage-dependent anion channel 2/3 (VDAC2/3)²⁰⁴; interestingly, VDAC2/3 expression was reduced in a NEDD4-dependent manner upon erastin treatment, thereby restricting erastin-induced ferroptosis¹⁶⁹. Other proteins with oncogenic activities, such as DJ1²⁰⁵ and pyruvate dehydrogenase kinase 4 (PDK4)²⁰⁶, can also promote ferroptosis evasion through corresponding mechanisms^145,149 (see section on targeting ferroptosis in cancer therapy). In all of these studies, inactivation of NEDD4, DJ1 or PDK4 promoted ferroptosis and enhanced the tumour-suppressive effects of ferroptosis inducers in xenograft models^145,149,169. Whether ferroptosis evasion supports the oncogenic roles of these proteins remains to be further explored.

Suppression of PUFA-PL synthesis and peroxidation.

Downregulation of peroxidised PUFA-PL levels in cancer cells has been linked to ferroptosis evasion and enhanced tumour growth. Calcium-independent phospholipase A₂β (iPLA₂β), a member of the iPLA₂ family, is overexpressed in some human cancers. iPLA₂β was recently shown to facilitate cancer cell escape from ferroptosis by hydrolysing peroxidised PLs, and iPLA₂β depletion sensitized cancer cells to ferroptosis and impaired xenograft tumour growth^90,91. Another recent study revealed that the expression of the adipokine chemerin is frequently upregulated in renal cell carcinoma (RCC), and that chemerin may maintain xenograft RCC growth through downregulating the levels of peroxidised PUFA-PLs and promoting ferroptosis evasion⁹².

Recent findings highlight that ferroptosis evasion through modulating fatty acid metabolism also contributes to cancer metastasis. Melanoma generally undergoes regional metastasis through the lymphatic system prior to systemic metastasis through the blood. The lymphatic environment has been observed to promote the escape of cancer cells from oxidative stress and ferroptosis in vivo, partly due to abundant levels of oleic acid (a MUFA) and GSH and low levels of free iron in the lymphatic fluid; oleic acid protects melanoma cells from ferroptosis in an ACSL3-dependent manner¹⁸ (as discussed in above, ACSL3-mediated MUFA-PL synthesis suppresses ferroptosis by displacing PUFAs from PLs). Importantly, ferroptosis resistance conferred by the lymphatic environment on cancer cells contributes to their subsequent survival during metastasis through the blood¹⁸. Likewise, under conditions of hypercholesterolemia, cancer cells exhibit superior metastatic capabilities in vivo, which may be attributable to ferroptosis resistance caused by increased accumulation of lipid droplets and MUFAs in these cells⁹³.

Restriction of the labile iron pool.

As an important cofactor in redox maintenance and iron homeostasis, iron–sulfur clusters (ISCs) and the associated regulatory pathways contribute to ferroptosis evasion in cancer cells by diminishing the labile iron pool⁹⁴. The cysteine desulfurase NFS1 is involved in ISC biosynthesis by harvesting sulfur from cysteine. NFS1 was found to be overexpressed in lung cancers and to be required for lung tumour growth in vivo, likely due to its ability to limit labile iron levels and protect cancer cells from ferroptosis⁹⁵. Frataxin, another protein involved in ISC assembly, is upregulated in different cancer types and was shown to promote cancer cell resistance to ferroptosis and, thus, xenograft tumour growth⁹⁶. Likewise, another recent study revealed that CDGSH iron-sulfur domain-containing protein 2 (CISD2), a component of the iron–sulfur protein family that is highly expressed in head and neck cancers, enhances the ability of cancer cells to counteract ferroptosis in vitro, likely by modulating ISCs and decreasing free iron levels⁹⁷. In addition, when breast cancer cells detach from the extracellular matrix, they can upregulate prominin 2 expression to promote the formation of ferritin-containing multivesicular bodies that export iron out of cells, thereby facilitating ferroptosis evasion in vitro⁹⁸. In some of the examples discussed above, whether ferroptosis evasion in cancer cells indeed contributes to enhanced tumour growth or metastasis in vivo remains to be established.

Upregulation of ferroptosis defences.

Most current studies of these mechanisms have centred on SLC7A11, which is overexpressed in multiple human cancer types^79,82. As discussed in Boxes 2 and 3, inactivation of tumour suppressors such as p53, BAP1, kelch-like ECH associated protein 1 (KEAP1) and ARF (p14), or activation of oncogenic KRAS, upregulates SLC7A11 expression, which can be dependent or independent of nuclear factor erythroid 2-related factor 2 (NRF2), conferring ferroptosis evasion and promoting tumour growth^{15,16,99–103}. As a master regulator of antioxidant defence, the transcription factor NRF2 governs the transcription of many genes involved in GPX4–GSH-mediated ferroptosis defence, including SLC7A11, thereby facilitating the escape of cancer cells from ferroptosis. Consequently, NRF2 signalling is upregulated in many human cancer types^99,101,104. Interestingly, a recent study showed that treatment with 27-hydroxycholesterol (27-HC) leads to downregulation of GPX4 levels in breast cancer cells, and that this effect is lost in breast cancer cells resistant to 27-HC, which likely endows such 27-HC-resistant breast cancer cells with increased resistance to ferroptosis and enhanced metastatic capability in vivo⁹³.

It appears that tumours also exploit other ferroptosis defence mechanisms. For example, GCH1 is overexpressed in multiple cancer types, including breast, lung, and liver cancers, and its high expression correlates with ferroptosis resistance in vitro⁷. FSP1- and DHODH-mediated ferroptosis defence centres on CoQ, whose synthesis is derived from the mevalonate pathway⁸³. Squalene monooxygenase (SQLE) is a key enzyme in the mevalonate pathway. A lack of SQLE expression in anaplastic lymphoma kinase (ALK)⁺ anaplastic large-cell lymphoma cells prevents cholesterol synthesis (thereby rendering such cancer cells cholesterol auxotrophic) and leads to the accumulation of the upstream metabolite squalene, a lipophilic antioxidant that protects cancer cells from ferroptosis in vivo¹⁰⁵. This mechanism provides a unique advantage for ALK⁺ anaplastic large-cell lymphoma growth under conditions of oxidative stress¹⁰⁵.

Ferroptosis vulnerability

Oncogenic mutations rewire cellular metabolic networks in cancer cells to fulfil their increased demand for nutrients and energy^106,107; this reprogramming often exposes new metabolic liabilities in cancer cells, rendering some of them uniquely vulnerable to ferroptosis. In this section, we conceptualize 3 such ferroptosis-based vulnerabilities in cancer cells, namely those induced by metabolic reprogramming, genetic mutations, and an imbalance in ferroptosis defences (Fig. 2a–c).

Figure 2. Ferroptosis as a vulnerability in cancer.

a. Therapy-resistant cancer cells with specific cellular states are vulnerable to ferroptosis, for example cancer cells with a mesenchymal phenotype that are enriched in polyunsaturated fatty acids (PUFAs) owing to high expression of zinc finger E-box binding homeobox 1 (ZEB1), elongation of very long-chain fatty acid protein 5 (ELOVL5) or fatty acid desaturase 1 (FADS1). Similarly, dedifferentiated subtypes of melanoma cells are characterized by PUFA accumulation and a deficiency of reduced glutathione (GSH), which render these cells vulnerable to ferroptosis. In addition, certain cancer cells, such as clear-cell renal cell carcinoma (ccRCC), non-neuroendocrine (NE) small-cell lung cancer (SCLC), and triple-negative breast cancer (TNBC) cells are inherently susceptible to ferroptosis owing to their unique metabolic features, such as high levels of PUFA-containing ether phospholipids (ePLs).

b. Mutations in certain tumour suppressors or oncogenes render cancers vulnerable to ferroptosis. Inactivating mutations in any constituent of the tumour-suppressive E-cadherin– neurofibromin 2 (NF2)–Hippo pathway confers a vulnerability to ferroptosis by upregulating Yes-associated protein (YAP)- or transcriptional coactivator with PDZ-binding motif (TAZ)-mediated transcription of ferroptosis-promoting factors, such as acyl-coenzyme A synthetase long chain family member 4 (ACSL4), transferrin receptor 1 (TfR1) and NADPH oxidase 4 (NOX4). In ccRCC, mutation or loss of Von Hippel-Lindau (VHL) promotes hypoxia inducible factor (HIF)-dependent expression of hypoxia-inducible, lipid droplet-associated protein (HILPDA), rendering ccRCCs vulnerable to ferroptosis. Non-small-cell lung cancers (NSCLCs) with epidermal growth factor receptor (EGFR) mutations are vulnerable to ferroptosis because of their high dependence on cystine. In another example, isocitrate dehydrogenase 1 (IDH1)-mutated cancer cells with increased levels of the oncometabolite 2-hydroxyglutarate (2-HG) are sensitive to ferroptosis owing to their decreased glutathione peroxidase 4 (GPX4) levels. * indicates either a mutation or loss of the gene, dependent on the gene.

c. Vulnerability to ferroptosis is triggered by an imbalance between GPX4-dependent and GPX4-independent ferroptosis defence systems. Cancer cells with low expression of components of GPX4-independent systems (such as ferroptosis suppressor protein-1 (FSP1), dihydroorotate dehydrogenase (DHODH) or GTP cyclohydroxylase-1 (GCH1)) depend on GPX4 for survival and therefore are vulnerable to GPX4 inhibition. Conversely, cancer cells with low expression of GPX4 are sensitive to inactivation of components of GPX4-independent systems. The dashed lines indicate that the function of the indicated protein is diminished or blocked in the corresponding context. PLs, phospholipids.

Metabolic features of cancer cells.

Several studies revealed that therapy-refractory cancer cells in specific cellular states are unexpectedly sensitive to ferroptosis^21,22,108. For example, cancer cells in a mesenchymal state, which are usually resistant to apoptosis induced by conventional therapies, are strongly dependent on GPX4, and this dependency is associated with high expression of zinc finger E-box binding homeobox 1 (ZEB1), a driver of epithelial-to-mesenchymal transition (EMT) and lipogenic factor¹⁰⁹. PUFA-PL synthesis is enhanced in cancer cells in the mesenchymal state, possibly owing to the central role of ZEB1 in lipid metabolism (Fig. 2a). The increased PUFA-PL levels in such cancer cells render them dependent on GPX4 to detoxify lipid peroxides for survival; consequently, these cells are highly vulnerable to ferroptosis²². Consistently, certain recurrent breast cancer cells with mesenchymal features exhibit upregulated discoidin domain-containing receptor 2 (DDR2) expression, which may be driven by EMT transcription factors; increased DDR2 promotes ferroptosis sensitivity in cancer cells through the Hippo pathway (the role of which in ferroptosis will be further discussed below)¹¹⁰. Likewise, key enzymes involved in PUFA synthesis, such as elongation of very long-chain fatty acid protein 5 (ELOVL5) and fatty acid desaturase 1 (FADS1), were found to be selectively overexpressed in mesenchymal gastric cancer cells, rendering these cancer cells particularly susceptible to ferroptosis¹¹¹ (Fig. 2a).

Similar to therapy-resistant mesenchymal-type cancer cells that depend on GPX4, drug-tolerant persister cancer cells also are highly sensitive to ferroptosis, perhaps because they also exhibit mesenchymal-like features²¹. Interestingly, CD44, a transmembrane glycoprotein that is highly expressed in mesenchymal cancer cells, mediates hyaluronate-dependent iron endocytosis¹¹², which might promote ferroptosis sensitivity in mesenchymal cancer cells through increasing cellular iron load. Moreover, therapy-resistant dedifferentiated subsets of melanoma cells also show a vulnerability to ferroptosis, possibly because of PUFA accumulation and low levels of GSH¹⁰⁸ (Fig. 2a). In sum, therapy-resistant cancer cells often exhibit altered metabolic states (such as increased PUFA-PLs associated with EMT) that confer a vulnerability to ferroptosis induction.

Certain cancer types are also inherently susceptible to ferroptosis owing to other unique metabolic features (Fig. 2a). As described in a previous section, PUFA-ePLs synthesized in peroxisomes provide substrates for lipid peroxidation²³. Clear-cell RCC (ccRCC) cells exhibit high levels of PUFA-ePLs, possibly owing to their high expression of alkylglycerone-phosphate synthase (AGPS, a key enzyme involved in PUFA-ePL synthesis), rendering these cells sensitive to ferroptosis²³. Within small-cell lung cancer (SCLC), non-neuroendocrine SCLC cells are much more sensitive to ferroptosis than are neuroendocrine SCLC cells, at least partly because non-neuroendocrine SCLC cells overexpress ePL synthesis enzymes and have high levels of PUFA-ePLs¹¹³. Finally, triple-negative breast cancer (TNBC) cells were found to be susceptible to ferroptosis. This susceptibility has been ascribed to several metabolic features of these cells, including their abundant PUFA levels, elevated labile iron pool and attenuated GPX4–GSH defence system^33,114. A vulnerability to ferroptosis uncovered in the abovementioned cancer types might provide new therapeutic opportunities for treating these highly intractable diseases.

Genetic mutations in cancer cells.

As discussed in Box 2, inactivation of tumour suppressors generally promotes ferroptosis resistance. However, in certain cases, tumour-suppressor mutations can confer an unexpected vulnerability to ferroptosis. The E-cadherin–neurofibromin 2 (NF2)–Hippo signalling axis presents a remarkable example of this type of vulnerability. Unlike other tumour suppressors (e.g., p53, BAP1, KEAP1 and ARF (p14), which promote ferroptosis; see Box 2), this tumour-suppressive axis suppresses ferroptosis²⁰. E-cadherin-mediated intercellular interactions activate Hippo tumour-suppressive signalling in an NF2-dependent manner, repress the transcriptional activity of Yes-associated protein (YAP) or transcriptional coactivator with PDZ-binding motif (TAZ) and downregulate the expression of several ferroptosis-promoting factors, thereby suppressing ferroptosis (which clearly does not align with the tumour-suppressor function of NF2)^20,115 (Fig. 2b). Consequently, inactivation of any component in the E-cadherin–NF2–Hippo pathway increases YAP or TAZ expression and/or activity, rendering cancer cells or tumours harbouring mutations in this pathway (such as NF2-mutant mesotheliomas) particularly susceptible to FINs^20,115. The Von Hippel-Lindau (VHL) tumour suppressor is another such example. VHL is lost in the majority of ccRCCs^116,117. Notably, VHL-deficient ccRCC cells are susceptible to ferroptosis, and restoration of wild-type VHL renders them insensitive to ferroptosis¹¹⁸. Several mechanisms might underlie this unique vulnerability in VHL-deficient ccRCCs, prominent among which is stabilization of hypoxia inducible factors (HIFs) induced by loss of VHL, which then promote PUFA synthesis through inducing the expression of hypoxia-inducible, lipid droplet-associated protein (HILPDA)^24,118 (Fig. 2b).

Oncogene activation can also render cancer cells susceptible to ferroptosis. Non-small-cell lung cancer (NSCLC) cells with epidermal growth factor receptor (EGFR) mutations are highly dependent on cystine and sensitive to ferroptosis induced by SLC7A11 inhibition or cystine deprivation¹¹⁹ (Fig. 2b). Interestingly, cystine promotes cell survival in EGFR-mutant cancer cells through GSH-independent mechanisms¹¹⁹, which is in line with recent data showing that cyst(e)ine promotes GPX4 protein synthesis and suppresses ferroptosis partly via GSH-independent mechanisms^119,120. As another example, isocitrate dehydrogenase 1 (IDH1)-mutated cancer cells exhibit increased levels of the oncometabolite 2-hydroxyglutarate (2-HG), which reduces GPX4 protein levels, sensitizing these cancer cells to ferroptosis¹²¹ (Fig. 2b). Collectively, it appears that oncogene activation and tumour-suppressor inactivation can either suppress or promote ferroptosis, depending on the context. Promotion of ferroptosis can create a vulnerability to ferroptosis for potential therapeutic targeting in corresponding cancer types.

Imbalanced ferroptosis defences.

Ferroptosis defence systems can be broadly divided into GPX4-dependent and GPX4-independent arms. Low expression or partial inactivation of one arm can render cancer cells highly dependent on the other for ferroptosis defence, and consequently highly susceptible to ferroptosis induced by the inactivation of the other defence arm (whereas normal cells with both arms intact might be unaffected by inhibition of one defence arm) (Fig. 2c). A therapeutic strategy targeting imbalanced ferroptosis defence would be similar to strategies targeting synthetic lethality; for instance, the use of poly(ADP-ribose) polymerase (PARP) inhibitors to treat BRCA1-deficient cancers, which is based on the concept that cells are dependent on 2 parallel DNA repair pathways, one involving PARP and the other requiring BRCA1¹²².

In support of this concept, analyses show that cancer cell lines with low expression of FSP1, DHODH or GCH1 are in general more vulnerable to GPX4 inhibitors than those with high expression of the corresponding gene (indeed, FSP1 was identified as the gene with the most striking such correlation in the Cancer Therapeutics Response Portal (CTRP))^5–9. Furthermore, GPX4 genetic ablation reduced tumour growth more potently in FSP1 knockout xenograft tumours than in FSP1 wild-type counterparts, and this reduction in tumour growth was attributed to ferroptosis induction⁵. Due to the unsuitability of current GPX4 inhibitors for in vivo studies (see further discussion in Perspective), whether this observation also applies to GPX4 inhibitor treatment of tumours remains to be established. Conversely, cancer cells with low expression of GPX4 are more dependent on FSP1 or DHODH for ferroptosis defence, and therefore are more sensitive to ferroptosis induced by inactivation of FSP1 or DHODH than are those with high GPX4 expression^5,6,9. DHODH inhibitors were found to suppress the growth of GPX4^low xenograft tumours more potently than that of GPX4^high ones⁹. Likewise, low expression of GPX4 (in concert with high expression of ALOX5) renders germinal centre B cell-like diffuse large B-cell lymphoma (DLBCL) cells vulnerable to ferroptosis induced by dimethyl fumarate (a compound approved for psoriasis treatment that depletes GSH through direct GSH succinylation)¹²³. These recent findings suggest that the expression levels of genes involved in ferroptosis defence can be explored as biomarkers to select patients for cancer treatment with FINs that target other ferroptosis defence arms.

Ferroptosis in the microenvironment

Recent studies have also revealed that the tumour microenvironment (TME), particularly its immune cells, dictates whether tumour-cell ferroptosis will occur. CD8⁺ cytotoxic T cells, the major executors of antitumour immunity in the TME, secrete interferon-γ (IFNγ), which subsequently inhibits cystine uptake in cancer cells through downregulating SLC7A11 expression, thereby augmenting lipid peroxidation and ferroptosis in tumours²⁸ (Fig. 3). Interestingly, IFNγ has also been shown to suppress SLC7A11-mediated cystine transport in macrophages¹²⁴, suggesting that IFNγ can regulate SLC7A11 expression and/or activity in both cancer and non-cancer contexts. Furthermore, immune checkpoint inhibitors (ICIs) and cyst(e)inase together potentiated T cell-mediated antitumour immune responses by synergistically promoting tumour ferroptosis, suggesting that ferroptosis is important in T cell-mediated antitumour activity and that blocking SLC7A11-mediated cystine uptake in combination with ICIs is a potential therapeutic strategy for cancer²⁸. Of note, ferroptotic cancer cells can release several immunostimulatory signals, such as high mobility group box 1 (HMGB1), calreticulin, ATP and PE, which promote dendritic cell maturation, increase the efficiency of macrophages in the phagocytosis of ferroptotic cancer cells and further enhance the infiltration of CD8⁺ T cells into tumours^125–128 (Fig. 3). Specifically, early ferroptotic cells (after 1- or 3-hours of treatment with a GPX4 inhibitor) by releasing immunostimulatory signals can promote the phenotypic maturation of dendritic cells and elicit a vaccination-like effect to activate antitumour immunity¹²⁷. Such evidence supports the concept that ferroptosis may act as a form of immunogenic cell death.

Figure 3. The role of ferroptosis in antitumour immunity.

Ferroptosis has a dual role to play in antitumour immunity, dependent on the nature of the immune cell. Boosting antitumour immunity, interferon-γ (IFNγ) secreted by CD8⁺ T cells promotes cancer cell ferroptosis by repressing solute carrier family 7 member 11(SLC7A11) expression in cancer cells. In turn, ferroptotic cancer cells release immunostimulatory signals that promote dendritic cell (DC) maturation and increase the efficiency of macrophages, particularly M1-like tumour-associated macrophages (TAMs), to phagocytose ferroptotic cancer cells. This further strengthens CD8⁺ T cell-mediated tumour suppression. In addition, several types of immunosuppressive cells, including regulatory T (T_reg) cells, myeloid-derived suppressor cells (MDSCs) and M2-like TAMs, are impaired by ferroptosis induction mediated by inhibition of glutathione peroxidase 4 (GPX4) or N-acylsphingosine amidohydrolase 2 (ASAH2), thereby augmenting antitumour immunity. However, CD8⁺ T cells and some T helper (T_H) cell subsets, such as T follicular helper (T_FH) cells, are also susceptible to ferroptosis, which compromises the contribution of ferroptosis to antitumour immunity. The dashed lines indicate that the function of the indicated cell or protein is diminished or blocked in the corresponding context. TCR, T cell receptor.

Ferroptosis induction in some immunosuppressive cells can also augment antitumour immunity (Fig. 3). Regulatory T (T_reg) cells, an immunosuppressive subset of CD4⁺ T cells that hinder protective immune surveillance against tumours, are relatively resistant to ferroptosis, possibly owing to GPX4 induction in activated T_reg cells^129,130. Correspondingly, T_reg cell–specific deletion of Gpx4 triggers ferroptosis in T_reg cells, contributing to antitumour immunity¹³⁰. Similarly, myeloid-derived suppressor cells (MDSCs) with immunosuppressive functions exhibit resistance to ferroptosis that is driven by N-acylsphingosine amidohydrolase 2 (ASAH2)-mediated inhibition of the p53–heme oxygenase 1 (HMOX1) axis¹³¹; consequently, targeting ASAH2 to induce ferroptosis in MDSCs increases the activation of tumour-infiltrating cytotoxic CD8⁺ T cells and promotes tumour suppression¹³¹. In addition, tumour-associated macrophages (TAMs) predominantly display the M2-like phenotype to suppress antitumour immunity¹³². Notably, immunosuppressive M2-like TAMs are more vulnerable to ferroptosis induced by inhibition of GPX4 owing to their lack of inducible nitric oxide synthase (iNOS) expression and nitric oxide (NO•) generation (NO• can detoxify lipid peroxides) than are the M1-like TAMs that promote antitumour immunity^126,133. Therefore, inducing ferroptosis in M2-like TAMs without affecting M1-like TAMs is a potential strategy to overcome the immunosuppressive TME and augment the effects of cancer immunotherapy¹³³.

However, emerging evidence also indicates that ferroptosis has a tumour-promoting effect in the context of tumour immunity. Gpx4-deficient T cells derived from T cell–specific Gpx4 knockout mice rapidly accumulate lipid peroxides in cell membranes upon T-cell activation and subsequently undergo ferroptosis¹³⁴. Furthermore, substantial lipid peroxides are detected in CD8⁺ T cells derived from tumours, but not those from lymph nodes, suggesting that ferroptosis may be a metabolic vulnerability of tumour-specific CD8⁺ T cells¹³⁵. Ferroptosis in this context would presumably dampen antitumour immunity and promote tumour growth. Supporting this idea, CD36 mediates fatty acid uptake by tumour-infiltrating CD8⁺ T cells in the TME, and increased CD36 expression induces lipid peroxidation and ferroptosis in CD8⁺ T cells, thereby compromising antitumour immunity¹³⁶. Genetic ablation of CD36 or blocking ferroptosis in CD8⁺ T cells effectively restores their antitumour activity, whereas CD8⁺ T cells treated with GPX4 inhibitors undergo ferroptosis and thus exert impaired antitumour effects in vivo¹³⁶. Furthermore, T follicular helper (T_FH) cells, a subset of CD4⁺ T cells that favours antitumour immunity, are highly susceptible to ferroptosis, and GPX4 expression is necessary for their survival and function¹³⁷ (Fig. 3). However, it should be noted that Slc7a11 knockout in mice or cystine deprivation in vivo does not decrease the viability or antitumour effects of T cells¹³⁸, which may explain the synergistic augmentation of antitumour immunity by targeting SLC7A11 in combination with ICIs²⁸. The mechanisms underlying the differential effects of GPX4 deletion versus SLC7A11 deletion on T cell function remain elusive, but might relate to the low expression (and likely a non-essential role) of SLC7A11 in T cells^139,140.

Targeting ferroptosis in cancer therapy

In recent years, FINs (Box 1) have garnered considerable interest in cancer research owing to their enormous therapeutic potential. Moreover, a variety of nanomaterials have been developed to induce ferroptosis locally or to enhance the activity of FINs³⁰. Furthermore, growing evidence suggests that ferroptosis at least partly mediates the tumour-suppressive effects of several conventional cancer therapies, including RT^25,141–143, chemotherapy²⁶, targeted therapy^27,144 and immunotherapy²⁸, and that FINs could potentiate the efficacy of these therapies by boosting tumour ferroptosis. As discussed below, inducing ferroptosis with FINs could be a promising therapeutic strategy to eliminate cancers with specific characteristics (Table 1).

Table 1. Strategies for targeting ferroptosis in cancer therapy.

27-HC, 27-hydroxycholesterol; ALK, anaplastic lymphoma kinase; BSO, buthionine sulphoximine; BH₄, tetrahydrobiopterin; ccRCC, clear cell renal cell carcinoma; CISD2, CDGSH iron-sulfur domain-containing protein 2; DHODH, dihydroorotate dehydrogenase; EGFR, epidermal growth factor receptor; EMT, epithelial-to-mesenchymal transition; FINs, ferroptosis inducers; FSP1, ferroptosis suppressor protein-1; GCH1, GTP cyclohydroxylase-1; GPX4, glutathione peroxidase 4; HMGCR, 3-hydroxy-3-methylglutaryl-coenzyme A reductase; ICIs, immune checkpoint inhibitors; IDH1, isocitrate dehydrogenase 1; iPLA₂β, calcium-independent phospholipase A₂β; KEAP1, kelch-like ECH associated protein 1; MT1G, metallothionein-1G; mTORC1, mTOR complex 1; NE, neuroendocrine; NCSLC, non-small-cell lung cancer; NEDD4, neural precursor cell expressed, developmentally down-regulated 4; NF2, neurofibromin 2; PD1, programmed cell death protein 1; PDK4, pyruvate dehydrogenase kinase 4; PDL1, PD1 ligand 1; PUFA-PL, polyunsaturated fatty acid-containing phospholipid; RCC, renal cell carcinoma; RT, radiotherapy; SCLC, small-cell lung cancer; SLC7A11, solute carrier family 7 member 11; SQLE, squalene monooxygenase; SQS, squalene synthase; TBH, tert-butyl-hydroperoxide; TNBC, triple-negative breast cancer; VHL, Von Hippel-Lindau.

Strategy	Classification	Feature	Tumour type	Target or agent
Exploiting the vulnerability of tumours to ferroptosis	Cancer cells with specific metabolic features	EMT	Multiple cancers	GPX4 inhibitors, HMGCR inhibitors22,111
		Drug-tolerant persister	Multiple cancers	GPX4 inhibitors21
		Dedifferentiation	Melanoma	SLC7A11 inhibitors, GPX4 inhibitors108
		ccRCC	RCC	GPX4 inhibitors23
		Non-NE SCLC	SCLC	SLC7A11 inhibitors, GPX4 inhibitors113
		TNBC	Breast cancer	SLC7A11 inhibitors, GPX4 inhibitors114
	Cancer cells with certain genetic mutations	Mutations in E-cadherin–NF2–Hippo axis	Multiple cancer types	SLC7A11 inhibitors, GPX4 inhibitors20,115
		VHL deficiency	RCC	GPX4 inhibitors, inhibition of glutathione synthesis (erastin, BSO)24,118
		EGFR mutation	NSCLC	Cysteine depletion119
		IDH1 mutation	Multiple cancer types	SLC7A11 inhibitors121
	Cancer cells with imbalanced ferroptosis defence	GPX4 low	Multiple cancer types	FSP1 inhibitors, DHODH inhibitors, GCH1 inhibitors, BH4 depletion5–9
		FSP1, DHODH or GCH1 low	Multiple cancer types	SLC7A11 inhibitors, GPX4 inhibitors5–9
Re-sensitizing resistant tumours to ferroptosis	Desuppression of PUFA-PL synthesis and peroxidation	iPLA2β overexpression	p53 wild-type cancers	iPLA2β inhibition + TBH90
		Adipokine chemerin upregulation	Renal cell carcinoma	Chemerin monoclonal antibody92
	Derestriction of the labile iron availability	NFS1 overexpression	Lung cancer	NFS1 inhibition + BSO, erastin or TBH95
		Frataxin overexpression	Multiple cancer types	Frataxin inhibition + erastin96
		CISD2 overexpression	Head and neck cancer	Pioglitazone (CISD2 inhibitor) + sulfasalazine (SLC7A11 inhibitor)97
		Prominin2 upregulation	Breast cancer	Prominin2 inhibition + SLC7A11 inhibitors or GPX4 inhibitors98
	Preventing the upregulation of ferroptosis defence systems	SLC7A11 upregulation	KRAS mutation	SLC7A11 inhibitors102,103
		GPX4 upregulation	Cancers chronically exposed to 27-HC	GPX4 inhibition93
		FSP1 or GCH1 high expression	Multiple cancer types	FSP1 inhibitors, GCH1 inhibitors, BH4 depletion5–8
		SQLE deficiency	ALK+ anaplastic large cell lymphoma	GPX4 inhibitors + inhibition of squalene synthesis (atorvastatin, zaragozic acid, SQS inhibition)105
	Abrogation of oncogenic activation-mediated ferroptosis resistance	PIK3CA activating mutations or PTEN deletion	Multiple cancer types	mTORC1 inhibitors + GPX4 inhibitors or SLC7A11 inhibitors17
		NEDD4 overexpression	Melanoma	NEDD4 inhibition + SLC7A11 inhibitors169
		DJ1 overexpression	Multiple cancer types	DJ1 inhibition + SLC7A11 inhibitors or GPX4 inhibitors145
		PDK4 overexpression	Pancreatic ductal adenocarcinoma	PDK4 inhibitors + SLC7A11 inhibitors149
Combining FINs with conventional cancer therapies	Sensitizing conventional cancer therapies	Chemotherapy, immunotherapy, or immunotherapy in combination with RT	Multiple cancer types	SLC7A11 inhibitors + cisplatin, doxorubicin, ICIs, or ICIs + RT26,28,143,153,154
		SLC7A11 or GPX4 induction upon RT	NSCLC	RT + SLC7A11 inhibitors or GPX4 inhibitors25
		GPX4 induction upon treatment with gemcitabine	Pancreatic ductal adenocarcinoma	Gemcitabine + SLC7A11 inhibitors152
	Overcoming resistance to conventional cancer therapies	Acquired radioresistance, intrinsic radioresistance (TP53 or KEAP1 mutation)	NSCLC	RT + SLC7A11 inhibitors or GPX4 inhibitors25,142,155
		Acquired chemoresistance (cisplatin or docetaxel)	Head and neck cancer, ovarian cancer	SLC7A11 inhibitors + cisplatin or docetaxel157,158
		TYRO3-mediated anti-PD1 or PDL1 therapy resistance	Breast cancer	TYRO3 inhibitors + anti-PD1159
		MT1G-mediated sorafenib resistance	Hepatocellular carcinoma	MT1G inhibitors + sorafenib160

Exploiting ferroptosis vulnerabilities.

As outlined in the preceding section, ferroptosis represents a vulnerability in certain cancer types, and targeting these vulnerabilities by inducing ferroptosis provides opportunities for cancer treatment. Notably, in several cancer types (such as lung and breast cancers), cancer cells appear to be more sensitive to ferroptosis than their corresponding normal epithelial cells in vitro¹⁴². These data highlight the existence of appropriate therapeutic windows that would allow selective ferroptosis induction in tumours while sparing normal tissues.

Re-sensitizing resistant tumours to ferroptosis.

As previously discussed, diverse ferroptosis escape mechanisms confer ferroptosis resistance in cancer cells. Disrupting the mechanisms that drive ferroptosis evasion can re-sensitize ferroptosis-resistant cancer cells or tumours to ferroptosis. Ferroptosis resistance mediated by certain genes with oncogenic activities (see Box 3) can be overcome by inhibiting expression or activity of the protein products of the genes themselves. For example, DJ1 depletion by disrupting the activity of S-adenosyl homocysteine hydrolase (SAHH) in the transsulfuration pathway re-sensitizes xenograft tumours to FINs that block SLC7A11¹⁴⁵. The transsulfuration pathway is constitutively active in conducting de novo cysteine synthesis in some cancer cells, whereas the expression of transsulfuration enzymes can be induced by cystine starvation in others^145–147. Therefore, while cancer cells mainly rely on SLC7A11-mediated cystine uptake to generate intracellular cysteine, the transsulfuration pathway still significantly contributes to the intracellular cysteine pool in some cancer cells, and protects them from cystine starvation-induced ferroptosis^146,147. Consequently, targeting enzymes involved in the transsulfuration pathway, such as the DJ1–SAHH axis, cystathionine β-synthase (CBS), and glycine N-methyltransferase (GNMT), can potentiate the susceptibility of tumours to FINs that block SLC7A11^145,146,148. As another example, pyruvate dehydrogenase kinase 4 (PDK4) inhibition by accelerating pyruvate oxidation and stimulating fatty acid synthesis can re-sensitize xenograft tumours to FINs that block SLC7A11¹⁴⁹.

Oncogene-induced ferroptosis resistance can also be mediated through downstream effectors that can be targeted to reverse ferroptosis resistance. For instance, PI3K promotes ferroptosis resistance mainly through mTOR complex 1 (mTORC1), and the combination of mTORC1 inhibitors with FINs that block SLC7A11 exerts potent tumour-suppressive effects in PIK3CA- or PTEN-mutant xenograft tumours¹⁷. Cysteine depletion by SLC7A11 inhibitors or cyst(e)inase exhibits striking efficacy against KRAS-mutant tumours, including those from genetically engineered mouse models, by impairing the enhanced ferroptosis defence (i.e., increased levels of intracellular cysteine and GSH) on which KRAS-mutant tumours depend^102,103,150. Overall, targeting key oncogenic pathways that confer ferroptosis resistance represents an important strategy for cancer therapy.

Combining FINs with conventional cancer therapies.

It is increasingly appreciated that diverse forms of conventional cancer therapy can trigger ferroptosis. Therefore, boosting ferroptosis induced by these therapies (for instance with FINs) could further strengthen their therapeutic efficacy. RT induces ferroptosis through multiple parallel mechanisms¹⁵¹. In response to RT, cancer cells evolve adaptive responses, such as upregulating SLC7A11 or GPX4 expression, to antagonize RT-induced ferroptosis²⁵. Consequently, FINs targeting either SLC7A11 or GPX4 in combination with RT can radiosensitise cancer cells or xenograft tumours by potentiating ferroptosis^25,141. Similarly, gemcitabine, a chemotherapeutic agent, induces GPX4 expression and activity; GPX4 inhibition counteracts this effect, increasing the sensitivity of cancer cells and xenograft tumours to gemcitabine by inducing ferroptosis¹⁵². Likewise, FINs targeting SLC7A11 can sensitise cancer cells to chemotherapy (e.g., cisplatin and doxorubicin)^153,154, immunotherapy (e.g., ICIs)²⁸, and immunotherapy in combination with RT¹⁴³.

Notably, for some cancer types with intrinsic or acquired therapy resistance, inducing ferroptosis with FINs could restore their sensitivity to conventional therapies. Mutations in the tumour suppressors TP53 or KEAP1 dampen RT-induced ferroptosis by upregulating SLC7A11 and other mechanisms, resulting in intrinsic radioresistance^25,142. Cancer cells with acquired radioresistance also exhibit SLC7A11 upregulation and resistance to ferroptosis^142,155,156. Consequently, reactivation of RT-induced ferroptosis by FINs can restore radiosensitivity in these cancer cells and xenograft tumours^142,155. Likewise, chemoresistance in certain cancer types can be overcome by FINs. It was reported that cisplatin resistance in head and neck squamous cell carcinoma cells and corresponding xenograft tumours could be abrogated by adding FINs targeting SLC7A11¹⁵⁷. In addition, FINs that inhibit SLC7A11 reversed docetaxel resistance in ovarian cancer cells in vitro¹⁵⁸.

In the examples discussed above, FINs were combined with conventional therapies to overcome therapy resistance. Likewise, specific inhibitors can be combined with conventional therapies to boost ferroptosis induction in tumours and reverse therapy resistance. Tumours with high TYRO3 expression are resistant to anti-programmed cell death protein 1 (PD1) or programmed cell death 1 ligand 1 (PDL1) therapy, while metallothionein-1G (MT1G) induction renders cancer cells resistant to the multikinase inhibitor sorafenib. It was recently revealed that therapeutic resistance induced by these proteins is partly due to ferroptosis blockade and can be overcome by restoring ferroptosis through combining anti-PD1 or PDL1 therapy with TYRO3 inhibitors or sorafenib with MT1G inhibitors in vivo^159,160.

Taken together, resistance to RT, chemotherapy, immunotherapy and other targeted therapies have been linked to ferroptosis resistance and this phenotype can be reversed by restoring ferroptosis with FINs or other inhibitors. It is important to note that the combination of FINs, especially class I FINs that inhibit SLC7A11, with conventional therapy appears to cause relatively limited toxic effects in normal cells¹⁴² and shows good tolerability in preclinical models^{25,28,141–143,159,160}, further supporting the clinical application of ferroptosis-inducing combination therapeutic strategies. Several clinical trials are currently under way to test the efficacy and safety of FINs or anticancer drugs with ferroptosis-inducing activity, either alone or in combination with conventional therapy, in patients with cancer (e.g., NCT04205357¹⁶¹, NCT04092647¹⁶², NCT02559778¹⁶³, NCT03247088¹⁶⁴).

Perspective

Ferroptosis is a form of metabolically regulated cell death. Metabolism in cancer cells is substantially rewired to meet their increased bioenergetic and biosynthetic needs and to support their rapid proliferation. This metabolic reprogramming often induces unique metabolic features, such as an enrichment of PUFA-PLs, an overload of iron and imbalanced ferroptosis defence systems, that can create a targetable vulnerability to ferroptosis, and represents an exciting opportunity for the discovery of new therapeutic targets in cancer. In addition, for tumours that exhibit intrinsic or acquired ferroptosis resistance, targeting the underlying resistance mechanisms can restore their vulnerability to ferroptosis. Therapeutic efficacy can be further enhanced by combining FINs with conventional therapies that can induce ferroptosis; importantly, such combination therapies have shown good synergism and tolerability in preclinical models. However, thorough histological and pharmacological analyses remain imperative to evaluate potential toxic effects of FINs in normal tissues and to determine the optimal drug dosing and scheduling in patients. To realize the full potential of ferroptosis-inducing strategies in cancer therapy, several additional challenges remain to be tackled in future investigations.

GPX4 constitutes the most powerful defence against ferroptosis, and several types of therapy-resistant tumours are remarkably vulnerable to GPX4 inhibition^22–24, highlighting the importance of GPX4 inhibitors in targeting ferroptosis vulnerability in cancer. The GPX4 covalent inhibitor ML210 and its recently designed derivatives (such as JKE-1674) are more selective than are RSL3 and ML162 (Box 1); however, most of these inhibitors have exhibited poor (or unclear) pharmacological properties in animal models, limiting their potential for clinical translation¹⁶⁵. Developing and optimizing GPX4-targeting agents with improved pharmacokinetics and selectivity remains a major barrier to employing GPX4 inhibition in cancer therapy. In this regard, several currently available and in vivo-stable anticancer agents with GPX4-inhibitory activity, such as withaferin A and altretamine, may offer an alternative to address this problem^166,167. In addition, considering that Gpx4 is an essential gene in the mouse^49,72, it also remains to be determined whether pharmacological inhibition of GPX4 can selectively kill tumours without inducing extensive toxicity in normal tissues and intolerable side effects in patients.

In addition to cancer cells, several cell types in the TME, including immune cells that promote or suppress antitumour immunity, might also be susceptible to ferroptosis. Therefore, how to balance the ferroptosis vulnerabilities of cancer cells, antitumour immune cells and immunosuppressive cells remains a critical obstacle. To address this issue, it will be important to comprehensively understand the mechanisms underlying the differential sensitivities of cancer cells and various immune cells to ferroptosis.

Furthermore, there remains an urgent need to develop predictive biomarkers that can accurately predict tumour responses to ferroptosis induction, especially those that can be tested directly in patients’ body fluids and biopsy specimens. Such tools will be essential for the stratification of patients with cancer for treatment with ferroptosis-inducing therapies.

More in-depth understanding of ferroptosis mechanism will continue to provide key insights into targeting ferroptosis in cancer. The recently proposed compartmentalization model⁹ suggests that there might exist additional ferroptosis defence systems localized in other organelles. Another key question in ferroptosis research is whether lipid peroxidation on the plasma membrane represents the final step or there are additional as-yet-unknown downstream steps in triggering plasma membrane rupture and ferroptotic cell death. The observation that cytosolic GPX4 (which can detoxify lipid peroxides accumulated on the plasma membrane) fails to suppress ferroptosis in DHODH/GPX4 double inactivated cells⁹ indicates additional ferroptosis execution mechanisms downstream of lipid peroxidation on the plasma membrane. Understanding these fundamental questions might identify new targets in cancer therapy.

Finally, it is believed that proteins involved in PUFA-PL synthesis and peroxidation (such as ACSL4 or POR) play a passive role in ferroptosis execution by providing a constitutive supply of PUFA-PLs or ROS for lipid peroxidation (in contrast, cell death executioner proteins involved in other cell death mechanisms, such as caspase, MLKL, and gasdermin D, play an active role in inducing cell death in response to physiological cues and, correspondingly, their activation is tightly controlled by upstream signalling). This view was challenged by a very recent study showing that initial ferroptotic stress activates protein kinase C β isoform 2 (PKCβII), which then phosphorylates ACSL4 and promotes ACSL4-mediated PUFA-PL synthesis and subsequent ferroptosis, thereby forming a positive feedforward loop to amplify ferroptosis¹⁶⁸. Attenuation of the PKCβII-ACSL4 signalling axis dampens the efficacy of immunotherapy by suppressing tumour ferroptosis¹⁶⁸. This study will inspire future investigations aimed to understanding additional regulatory mechanisms involved in ferroptosis execution and their relevance to cancer therapy. We envision that the next few years will witness exciting new findings that will translate our mechanistic understanding of this intriguing cell death mechanism into effective cancer therapies.

Glossary

Cristae

Folds of the inner mitochondrial membrane that extend into the matrix of a mitochondrion.

Ferroptosis inducers

(FINs). A compound or treatment that can induce ferroptosis by boosting ferroptosis-promoting mechanisms and/or suppressing ferroptosis defence mechanisms.

Polyunsaturated fatty acid

(PUFA). A fatty acid that contains more than one double bond and is required for cellular signalling and membrane fluidity.

Peroxisomes

Organelles that are important for β-oxidation of very-long-chain fatty acids and synthesis of ether phospholipids.

Fenton reaction

A nonenzymatic reaction of labile iron and hydrogen peroxide (H₂O₂) that generates hydroxide and hydroxyl radicals, which can subsequently induce lipid peroxidation.

Anaplerotic reactions

Metabolic reactions that replenish the supply of intermediates involved in the citric acid cycle.

GPX

(Glutathione peroxidase). A family of peroxidases that use reduced glutathione as their cofactor to reduce hydroperoxide species to their corresponding alcohols.

System x_c⁻

An antiporter that imports cystine and exports glutamate; it consists of 2 subunits, including the transporter subunit SLC7A11 and the regulatory subunit SLC3A2.

Ubiquinone

(coenzyme Q or CoQ). A lipophilic molecule that is composed of a quinone head group linked to a polyisoprenoid lipid tail and acts as an electron transport carrier in mitochondria.

Ubiquinol

(CoQH₂). The fully reduced form of ubiquinone.

Dihydroorotate dehydrogenase

(DHODH). An inner mitochondrial membrane-localized enzyme that oxidizes dihydroorotate to orotate for pyrimidine synthesis while reducing CoQ to CoQH₂

Hypercholesterolemia

High levels of cholesterol in the blood.

Lipid droplets

Organelles with a phospholipid monolayer that are responsible for lipid storage, including PUFA storage.

Iron–sulfur clusters

Molecular ensembles of iron and sulfur that function as protein co-factors to regulate iron homeostasis and redox reactions in response to oxidative stress.

Mevalonate pathway

A metabolic pathway that synthesizes cholesterol, CoQ and steroid hormones.

Epithelial-to-mesenchymal transition

(EMT). A process by which epithelial cells gradually lose their cell polarity and intercellular adhesion properties and acquire mesenchymal-like phenotypes including migratory and invasive properties.

Cyst(e)inase

Engineered enzymes that degrade extracellular cysteine and cystine.

Transsulfuration pathway

A metabolic pathway that transfers sulphur from homocysteine to cysteine, leading to cysteine biosynthesis.