Mechanisms of ferroptosis sensitization and resistance

Abstract

Ferroptosis is an iron-dependent and oxidative form of non-apoptotic cell death with roles in development, homeostasis, and disease. Ferroptosis sensitivity can vary between cells, often for reasons that are not well understood. In this Perspective, we describe the core ferroptosis mechanism and outline how changes in iron, redox, and lipid metabolism can alter ferroptosis sensitivity. We propose the concept of a ferroptosis sensitivity continuum to describe how different intrinsic and extrinsic factors interact to push cells towards a more ferroptosis-sensitive or ferroptosis-resistant state, with effects on development and diseases such as cancer.

INTRODUCTION

Ferroptosis is an oxidative form of non-apoptotic cell death.¹ This process is defined by the iron-dependent accumulation of membrane lipid peroxides.² Excessive membrane lipid peroxidation results in plasma membrane stiffening³ that opens ion channels and leads to osmotic swelling and eventual cell rupture. These processes are accompanied by the release of diffusible factors that can propagate a “wave” of ferroptosis between neighboring cells.^4–6 Coordinated ferroptotic waves help sculpt developing muscle tissue in the embryonic chick limb.⁷ However, this process must be tightly controlled as mutant studies indicate that aberrant ferroptosis is a threat to normal mammalian embryonic development.^8,9 In vivo, acute or chronic insults to the kidney, heart, brain, and other tissues can cause pathological ferroptosis.^5,10–13 Diverse animals and plants exposed to changes in diet or environmental stresses can also undergo ferroptosis or ferroptosis-like cell death.^14–17 Collectively, the investigation of ferroptosis is of broad biomedical interest.

In this Perspective, we tackle a fundamental question: What makes a cell sensitive or resistant to ferroptosis? This knowledge is important if we are to understand why one cell (but not another) undergoes ferroptosis in a developing tissue, in response to stress, or in response to drug treatment. Here, we describe the ferroptosis mechanism and discuss how and why ferroptosis sensitivity varies by cell type, cell state, and in response to changes in the environment. We introduce the concept of a ferroptosis sensitivity continuum to help understand the impact of these factors. Throughout, we will highlight important unanswered questions.

THE CORE FERROPTOSIS MECHANISM

Ferroptosis is caused by an imbalance between the accumulation of membrane lipid peroxides and the neutralization of these species (Figure 1). Membrane lipid peroxides are formed through chemical reactions between iron, oxygen, and “oxidizable” lipids.¹⁸ Membrane phospholipids containing polyunsaturated fatty acids (PL-PUFAs) are highly oxidizable and are key substrates required for the execution of ferroptosis (Figure 1A).¹⁹ Iron-containing lipoxygenases (LOX) may initiate ferroptosis in some cells by directly catalyzing PL-PUFA peroxidation.^20–23 However, enzyme-catalyzed lipid peroxidation is not absolutely required for ferroptosis.²² PUFA acyl chains contain hydrogen atoms that can be abstracted by other radical species (Figure 1B). The resulting carbon-centered radicals can combine with O₂ to yield a lipidperoxyl radical (L-OO^•). Further reactions between one lipidperoxyl radical and a neighboring PUFA acyl chain (to yield L-OOH and another carbon-centered radical) and between L-OOH and iron (to yield L-O^•) can result in the propagation of lipid peroxidation reactions throughout a membrane (Figure 1C).¹⁸ Iron chelation inhibits ferroptosis by preventing this chemical reaction.

Figure 1. Ferroptosis prerequisites and defense mechanisms.

Ferroptosis is caused by the accumulation of membrane lipid peroxides beyond the capacity of the cell to detoxify them. It is impossible to depict all chemical reactions, metabolites and proteins known to regulate ferroptosis sensitivity and resistance in one diagram. The above scheme outlines only some of the key components. (A) Membrane lipid peroxidation leading to ferroptosis requires oxidizable phospholipids (PLs). Oxidizable phospholipids contain polyunsaturated fatty acids (PUFAs). Certain PUFA free fatty acids are activated into PUFA-CoAs by ACSL4, then incorporated into membrane lipids by lysophospholipid acyltransferases like LPCAT3. The sequestration of PUFAs in neutral lipids (inside lipid droplets) can alter ferroptosis sensitivity. PUFA-containing phospholipids can be oxidized by lipoxygenase enzymes. (B) Lipoxygenase (LOX) enzymes can directly oxidize PUFA-containing phospholipids. Hydrogen atoms may also be abstracted from PUFA-containing phospholipids by reaction with one or more oxidizing species generated by NADPH oxidase (NOX), or by enzymes in the endoplasmic reticulum or mitochondria. (C) Intracellular labile iron can participate in the propagation of lipid peroxidation by reacting with lipid hydroperoxides to generate lipid radicals. The import and storage of iron therefore affects ferroptosis sensitivity. (D) Cells have several defense mechanisms against lipid peroxidation. Cystine is imported into the cell by system x_c⁻ and used to synthesize glutathione (GSH), a co-factor for the phospholipid hydroperoxidase glutathione peroxidase 4 (GPX4). (E) Lipid radicals (not directly depicted) can be neutralized by products of the FSP1-CoQ₁₀-vitamin K system, where FSP1 reduces CoQ₁₀ and vitamin K into their reduced forms. (F) Lipid radicals can also be neutralized by the GTP-GCH1-BH2/4 system, where GTP is used by GCH1 to synthesize BH2, which can then be reduced by dihydrofolate reductase (DHFR) to generate the potent radical trap tetrahydrobiopterin (BH4). DMT1, divalent metal transporter 1 (gene name: SLC11A2). NMTs, N-myristoyltransferases; PTS, 6-pyruvoyl-tetrahydropterin synthase; SR, sepiapterin reductase (gene name: SPR).

During ferroptosis, membrane lipid peroxides initially accumulate inside the cell on the membranes of organelles, including lysosomes and the endoplasmic reticulum, before accumulating at the plasma membrane prior to lethal permeabilization.^24–26 Plasma membrane lipid peroxidation is necessary for ferroptosis.^24,27 The accumulation of plasma membrane lipid peroxides triggers a change in membrane stiffness and permeability that leads ineluctably to membrane permeabilization. The death of one cell by ferroptosis can, in turn, trigger the death of neighboring cells in a wave-like manner.^4,5 These waves can travel at a constant speed (~5 μm/min) over a distance of several millimeters and are likely propagated by the release of redox and lipid mediators from one damaged cell to another.^6,7

In cells, the threat of overwhelming membrane lipid peroxidation is mitigated by several enzymes and metabolites. Glutathione peroxidase 4 (GPX4) is a key anti-ferroptotic protein in culture and in vivo.^28–30 GPX4 uses reduced glutathione (GSH) as a co-substrate to convert pro-ferroptotic lipid hydroperoxides (L-OOH) into non-toxic lipid alcohols (L-OH). Upstream of GPX4, the system x_c⁻ cystine/glutamate antiporter supplies cells with cystine required for the synthesis of GSH (Figure 1D).³¹ Cystine may also be used to synthesize coenzyme A, which can contribute to ferroptosis suppression in a manner that remains to be clarified.^32,33 In addition to the cystine-GSH-GPX4 axis, the cell is also protected from membrane lipid peroxidation by the ferroptosis suppressor protein 1 (FSP1)-coenzyme Q₁₀ (CoQ₁₀)-vitamin K system (Figure 1E)^34–36 and the GTP cyclohydrolase 1 (GCH1)-tetrahydrobiopterin (BH2/4) system (Figure 1F).^37,38 Hydropersulfides (e.g., RSSH) can directly inhibit lipid peroxidation reactions.^39,40 Other metabolites like 7-dehydrocholesterol (7-DHC)^41,42, vitamin E⁴³, and retinol^44,45 can also directly trap lipid radicals and thereby prevent overwhelming lipid peroxidation and ferroptosis.

To induce ferroptosis, these protective barriers must be overcome. Synthetic small molecules can inhibit the function of key barriers, thereby inducing ferroptosis. For example, erastin and RSL3 can induce ferroptosis by inhibiting system x_c⁻ and GPX4, respectively. These treatments are sufficient to induce ferroptosis in many but not all cells.^28,31 System x_c⁻ may also be inhibited by high concentrations of extracellular glutamate, a more physiological stimulus that may be relevant in brain diseases.⁴⁶ Synthetic inhibitors of FSP1 can induce ferroptosis in some cancer cells alone, and in other cancer cells, they can synergize with GPX4 inhibitors.^47,48 Anti-ferroptosis barriers can also be overcome by incubating cells with high concentrations of highly oxidizable conjugated PUFAs^49,50, iron-containing nanoparticles⁵¹, or iron itself.⁵² In some models, cold stress appears sufficient to stimulate lipid peroxidation and induce ferroptosis.⁵³ The ability of these different stimuli to induce ferroptosis despite the plethora of redundant defense mechanisms is remarkable and not yet fully understood. Moreover, while GPX4 is often the most important anti-ferroptotic protein in most cells^11,28, it is difficult to predict which other defense mechanism(s) may be important cell to inhibit ferroptosis in a given cell.

UNIQUE FEATURES OF THE FERROPTOSIS MECHANISM

Ferroptosis is unique compared to other forms of cell death.⁵⁴ Not only is this mechanism governed by distinct molecular regulators, but many different inputs can render a particular cell more ferroptosis-sensitive or ferroptosis-resistant, and these inputs can vary considerably between cells and systems. Appreciating the distinct nature of ferroptosis is important when thinking about how this process may be exploited naturally or in therapeutic settings. Below we highlight several inherent features of the ferroptosis mechanism that require consideration when thinking about this mechanism.

Ferroptosis emerges from normal cellular functions

Ferroptosis does not appear to require any single, dedicated death-promoting protein.⁵⁴ Rather, ferroptosis emerges from chemical interactions between atoms and molecules which otherwise support normal cellular functions: iron is an essential enzyme cofactor, oxygen is the terminal electron acceptor in mitochondrial ATP synthesis, and PUFA-containing lipids help maintain membrane fluidity and are needed for cell signaling, among other functions. Importantly, the mere presence of iron, oxygen, and PUFA-containing lipids in a cell does not indicate whether a cell is undergoing or likely to undergo ferroptosis, and high or low levels of these pro-ferroptotic factors may simply reflect the normal physiology of the cell.

Together, iron and other pro- and anti-ferroptotic molecules establish a latent cell state that is more or less conducive to the execution of ferroptosis, should the right conditions arise. In principle, any of the thousands of different molecules in a cell (e.g., RNAs, transcription factors, transporters, enzymes, metabolites) that alter the abundance of iron, redox metabolites, or lipids, have the chance to quantitatively influence ferroptosis sensitivity.⁵⁵ These thousands of different molecules have other functions in the cell, but through their normal functions, they create a context that is more or less favorable to the execution of ferroptosis. For example, the ferritinophagy adaptor protein nuclear receptor coactivator 4 (NCOA4) promotes the homeostatic catabolism of iron-loaded ferritin molecules in the lysosome. This process is needed for proper iron handling, especially in erythrocytes.^56,57 However, by increasing the levels of labile intracellular iron available to react with lipid hydroperoxides, NCOA4-mediated ferritin catabolism can heighten ferroptosis sensitivity.^58–60 Conceptually similar, the lipid metabolic enzyme lysophosphatidylcholine acyltransferase 1 (LPCAT1) is essential for the synthesis of phospholipids containing the saturated fatty acid palmitate.^61,62 However, by promoting the incorporation of these less oxidizable saturated fatty acids in place of more oxidizable PUFAs into membrane phospholipids, LPCAT1 overexpression in cancer cells can reduce ferroptosis sensitivity and, in turn, enhance tumor growth.⁶³ In this example, cancer cells appear to have evolved a mechanism to exploit the normal function of an enzyme in lipid metabolism to generate a ferroptosis-resistant state. On the other hand, some adaptations that cancer cells acquire to enhance their own fitness during the process of metastasis, such as increases in PUFA metabolism, may in turn create a heightened vulnerability to ferroptosis⁶⁴. Apparent differences in ferroptosis sensitivity between normal and diseased cells resulting from these changes do not become apparent (i.e., “revealed”) until the cell experiences a pro-ferroptotic stress.

Lipid peroxidation itself is not even unusual: some amount of lipid peroxidation is normally present in mammalian cells⁶⁵ and contributes to cell signaling and homeostasis.^66,67 Crucially, membrane lipid peroxide accumulation is typically limited. Lipid peroxidation only leads to ferroptosis when it exceeds some poorly defined threshold. We do not know exactly how much, what proportion of, or how quickly PL-PUFAs need to be peroxidized to trigger ferroptosis. We also do not know if or how this varies between cells. These unknowns make it difficult to predict ferroptosis sensitivity a priori from the levels of any one molecule. That said, differences in global lipid peroxidation susceptibility do correlate with differences in ferroptosis sensitivity.⁶⁸ This is consistent with the notion that the entire metabolic ensemble of the cell is integrated to dictate ferroptosis sensitivity.

Ferroptosis execution is contingent on diverse conditions being met

Ferroptosis occurs if the array of factors needed to accumulate membrane lipid peroxides to toxic levels are present and protective anti-ferroptotic barriers are missing, overcome, or inactivated. For instance, iron chelation alone is sufficient to completely shut down ferroptosis in an otherwise sensitive cell.^69,70 Certain amino acids, such as glutamine and arginine, must also be present for certain cancer cells to execute ferroptosis, though the underlying mechanisms need further clarification.^71,72 Also, high levels of PL-PUFAs are crucial for ferroptosis. Ferroptosis can be inhibited by incubating cells with small amounts of less oxidizable monounsaturated fatty acids (MUFAs).^15,23,24,73 MUFAs can take the place of PUFAs in the same phospholipid species, either during de novo synthesis or during the process of acyl chain remodeling.⁷⁴ These processes require the activation of MUFA free fatty acids to MUFA-CoAs by acyl-CoA synthetase long chain family member 3 (ACSL3, Figure 1A). Evidence from cultured fibrosarcoma cells and metastatic melanoma models in mice indicates that ACSL3-dependent MUFA metabolism promotes ferroptosis resistance in culture and in vivo.^24,73,75 Downstream of ACSL3, the lysophospholipid acyltransferases membrane bound O-acyltransferase domain containing (MBOAT) 1 and 2 are especially important for inserting MUFA-CoA into lysophospholipid backbones to prevent ferroptosis (Figure 1A).^74,76 Simply elevating MUFA levels can render GPX4 completely dispensable for cell survival, presumably by sufficiently reducing the “oxidizability” of membrane phospholipids to a point where overwhelming lipid peroxidation is no longer a threat.^24,73 Beyond these examples, a full accounting of all the factors that must be present or absent for ferroptosis to occur in a given cell has yet to be made.

Ferroptosis is mechanistically flexible

There is likely no universal ferroptosis regulatory pathway.⁵⁵ In different cells, the molecules or pathways that execute or critically regulate ferroptosis appear to vary. For example, depending on the cell type, the lipids most important for the execution ferroptosis may be PUFA-containing phosphatidylethanolamines¹⁹, PUFA-containing ether lipids^77,78, rare phosphatidylethanolamines and phosphatidylcholines containing two PUFA acyl chains^79,80, PUFA-containing phosphatidylinositols⁸¹, or PUFA-containing phosphatidylserines.⁸² Different lipid metabolic pathways may also be more or less important for ferroptosis depending on the cellular context. In renal cell carcinomas, the ether lipid synthesis pathway appears to be essential for ferroptosis.⁷⁷ By contrast, in fibrosarcoma cells, this pathway appears to be dispensable for ferroptosis execution²⁷, and in C. elegans, ether lipid synthesis may even promote ferroptosis resistance.⁸³

Different enzymes can even be more or less important for ferroptosis in the same cell, depending on how this process is induced. Acyl-CoA synthetase long chain family member 4 (ACSL4) provides a canonical example. ACSL4 activates certain PUFA free fatty acids to PUFA-CoAs (Figure 1A). PUFA activation is necessary for these species to be incorporated into membrane phospholipids.^19,80,81 ACSL4 is crucial for the execution of ferroptosis when this process is triggered by direct inhibition of GPX4.^84,85 On the other hand, ACSL4 seems dispensable when ferroptosis is triggered by cystine deprivation²⁷ or by other ferroptosis-inducing stimuli like tert-butyl hydroperoxide²¹, expression of pleckstrin homology-like domain family A member 2 (PHLDA2)⁸⁶, or photodynamic therapy⁸⁷. This indicates that different lipids and lipid metabolic enzymes are more or less important for the execution of ferroptosis depending on how this process is triggered.

Similar complexity is observed for anti-ferroptotic systems. GPX4 is essential for normal development in mice and humans.^30,88 While evading other forms of cell death, cancer cells often become more dependent on GPX4 for survival compared to normal cells.^89–91 Nonetheless, GPX4 is not required to prevent ferroptosis in all cells. In cancer cells expressing high levels of FSP1, GPX4 inhibition can be insufficient to induce ferroptosis.^34,47 In other cancer cells, the metabolites squalene, 7-DHC, and BH4 can accumulate to high levels and likewise compensate to varying degrees for the loss of GPX4 function.^{37,38,41,42,92} Thus, the relative importance of individual molecules for ferroptosis regulation can vary considerably between cells.

Finally, some metabolites, like NADPH, appear to contribute to both ferroptosis execution (as a substrate for NADPH oxidases that synthesize superoxide, Figure 1B) and ferroptosis resistance (as a co-substrate for FSP1, Figure 1E).⁵⁵ The sensitivity of a given cell to ferroptosis reflects the underlying cell state created by these and other molecules that increase or decrease the likelihood that the plasma membrane will be overwhelmed by lipid peroxide accumulation.⁶⁸ Predicting which metabolites, proteins, or pathways are critical for regulating ferroptosis, and how they might function in a given context, can be difficult. This challenge is exacerbated by the flexible regulation of ferroptosis sensitivity by many different mechanisms and factors, as described next.

STABLE DIFFERENCES IN FERROPTOSIS SENSITIVITY

Different mammalian cells appear inherently more or less ferroptosis sensitive. While embryonic Gpx4 deletion is lethal⁹³, tissue-specific Gpx4 deletion and whole-body hypomorphic Gpx4 mutations yield examples of cell-specific differences in ferroptosis sensitivity. When Gpx4 is selectively disrupted in the neurons of adult mice, spinal cord motor neurons degenerate within a week without concurrent loss of cortical neurons.⁹⁴ By contrast, when Gpx4 activity is globally reduced in mice via a selenocysteine to cysteine mutation in its active site, it is parvalbumin-positive GABAergic interneurons in the brain that degenerate first.²⁹ At the organismal level, inducible deletion of Gpx4 in adult mice leads to rapid kidney failure, with tubular epithelial cells especially affected¹¹. It is still unclear which molecular differences render these cells so sensitive to ferroptosis. Interestingly, mouse kidney tubule epithelial cells normally have high Gpx4 levels.¹¹ Perhaps these cells normally encounter a severe lipid peroxidation threat due to their iron, redox, and/or lipid state, making them especially sensitive to Gpx4 loss. A caveat to these studies is that incomplete Gpx4 genetic disruption, or differences in the rates of Gpx4 protein loss following gene disruption, could also lead to the apparent differences in ferroptosis sensitivity between cells and tissues in vivo.

Small molecule ferroptosis inducers provide a complementary way to examine stable differences in ferroptosis sensitivity. These small molecules can be applied to panels of cultured cells. In cultured cancer cells, consistent differences in ferroptosis sensitivity are observed. Diffuse large B-cell lymphoma (DLBCL) cells are more sensitive to the system x_c⁻ inhibitor erastin than cells from other blood cancer lineages, including acute myeloid leukemia and multiple myeloma.^28,95 Diving deeper, different sub-types of DLBCL appear to have unique sensitivities to GPX4 inhibition due, in part, to differences in ACSL4 function.⁹⁶ In the NCI-60 cancer cell line panel, kidney cancer cells are more sensitive to erastin than lung, colon, CNS, or other cancer cells.²⁸ Similarly, from a panel of several hundred cancer cell lines, clear cell carcinomas (CCCs) of the kidney and ovary are, on average, the most sensitive to the GPX4 inhibitor ML210, whereas pancreatic, urinary, and esophageal cell lines are least sensitive⁹⁷. Mechanistically, the high sensitivity of renal CCCs to ferroptosis may be driven by high relative PUFA-phospholipid abundance.

Cancer cells from mesenchymal lineages, such as sarcomas, are generally more sensitive to GPX4 inhibition than cancer cells from epithelial lineages.⁸⁹ Indeed, cancer cell lines of epithelial origin can sometimes tolerate GPX4 genetic disruption and continue to proliferate normally, whereas those of mesenchymal origin cannot.⁸⁹ Mechanistically, mesenchymal cells typically express higher levels of the transcription factor zinc finger E-box binding homeobox 1 (ZEB1). ZEB1 can alter the expression of other genes, which ultimately leads to the increased proportion of oxidizable PUFAs in the membrane at the expense of MUFAs.⁹⁸ This may provide a mechanistic explanation for the increased ferroptosis sensitivity in these cells: cancer cells with higher levels of PUFA-containing versus MUFA-containing lipids require GPX4 to protect against lipid peroxidation, whereas those with lower levels do not. Adopting this insight more broadly, a reasonable speculation is that those cells in normal mice that are more sensitive to Gpx4 loss have higher PL-PUFA/PL-MUFA ratios.

Sex-associated differences impact ferroptosis sensitivity in consistent ways. For example, in proximal kidney tubule cells, sensitivity to Gpx4 loss is sexually dimorphic: cells from females are more resistant to ferroptosis compared to those from males.⁹⁹ This is explained by higher expression of the “antioxidant” transcription factor NFE2 like bZIP transcription factor 2 (NFE2L2, NRF2) in female versus male cells. Intact ovaries are required for this ferroptosis protection, suggesting a hormonal effect. However, the precise mechanism linking ovary function to NFE2L2 expression remains to be fully elucidated, and other mechanisms of protection in female kidney cells are conceivable. In another example, sex hormone signaling through the estrogen or androgen receptors can elevate MBOAT1 or MBOAT2 levels, which in turn, enhances MUFA metabolism and reduces overall ferroptosis sensitivity in breast and prostate cancer, respectively (Figure 1A).⁷⁴ These differences will presumably contribute to differences in ferroptosis sensitivity between cells and tissues in a sex-specific manner.

Finally, differences in genotype can stably alter ferroptosis sensitivity. This is commonly observed in cancer and can involve many different genes. For example, a rare p53 allele (P47S), common to individuals of African descent, rewires intracellular metabolism to promote ferroptosis resistance and increase spontaneous tumor formation.¹⁰⁰ The p53 P47S mutation is associated with reduced transactivation of glutaminase 2 (GLS2). GLS2 converts glutamine to glutamate. The synthesis or accumulation of glutamate is associated with the enhanced production of oxidizing species that initiate ferroptosis.^72,101,102 At the same time, the p53 P47S mutation also increases the synthesis of GSH and CoA which can suppress ferroptosis.^32,33 In another example, KEAP1 mutations in lung cancer result in the aberrant accumulation of NFE2L2/NRF2 and enhanced transcription of the downstream target gene, AIFM2 (encoding FSP1), resulting in increased ferroptosis resistance in mice.¹⁰³ In glioblastoma, CDKN2A deletions, found in a subset of patient tumors, are associated with heightened sensitivity to GPX4 inhibition and ferroptosis.¹⁰⁴ This is explained by the re-routing of PUFAs from the triglyceride (TAG)/lipid droplet (LD) pool into the plasma membrane phospholipid pool. In breast and ovarian cancers, BRCA1 loss can promote resistance to erastin-induced ferroptosis in cultured cancer cells. However, this same mutation can increase ferroptosis sensitivity in response to GPX4 inhibition¹⁰⁵, illustrating once again how differences in ferroptosis sensitivity can vary depending on the trigger.

DYNAMIC REGULATION OF FERROPTOSIS SENSITIVITY

Ferroptosis sensitivity can be potently altered by dynamic changes in the environment of the cell and the cell state. These changes can occur in cultured cells over a matter of hours, days, or weeks. The different categories outlined below may overlap with each other and are not exhaustive. Still, what follows provides a framework for discussion of the highly plastic nature of ferroptosis sensitivity.

Extracellular metabolic milieu

The composition of the extracellular milieu has a significant impact on ferroptosis sensitivity. GPX4 is a selenoenzyme and cells require selenium uptake to support GPX4 synthesis.¹⁰⁶ Selenium concentration varies between batches/lots of serum, which can affect the synthesis of GPX4 protein in cultured cells.¹⁰⁷ Specialized growth medium containing physiological levels of sodium selenite can better prevent lipid peroxidation by supporting GPX4 expression.¹⁰⁸ Albumin is a cysteine-rich protein that is highly abundant in serum. When albumin is internalized and catabolized in the lysosome, this can provide cells with a source of cysteine to protect against ferroptosis when extracellular cystine is limiting.¹⁰⁹ The lipid composition of the environment is also important. Increased uptake of extracellular PUFAs can shift cells from a ferroptosis-resistant to a ferroptosis-sensitive state.¹¹⁰ By contrast, increased levels of extracellular MUFAs are sufficient to shift cells towards a ferroptosis-resistant state that can tolerate even with the complete inactivation of GPX4.²⁴ This has important functional consequences for cancer cells in vivo. Lymph fluid is a MUFA-enriched microenvironment; cancer cells that metastasize through the lymphatic system can take up these MUFAs and become more resistant to ferroptosis compared to those that metastasize through the bloodstream.⁷³ Finally, the pH of the environment influences ferroptosis sensitivity. An acidic tumor microenvironment can promote ferroptosis by enhancing the peroxidation of membrane lipids.¹¹¹ In sum, ferroptosis sensitivity is highly dependent on the metabolic milieu surrounding the cell.

Proliferative state

Depending on the context, arresting the proliferation of cancer cells can either increase or decrease ferroptosis sensitivity. Stabilizing the expression of wild-type p53 results in gene expression changes (e.g., induction of CDKN1A) that inhibit cell cycle progression. Wild-type p53 stabilization can increase resistance to ferroptosis triggered by cystine deprivation.^112,113 This mechanism of resistance involves the downregulation of nucleotide synthesis and the conservation of intracellular GSH. However, p53-stabilized cells can be more resistant to ferroptosis caused by cystine deprivation and at the same time more sensitive to ferroptosis triggered by direct GPX4 inhibitors.⁷⁶ The mechanism of sensitization to GPX4 inhibition involves the transcriptional downregulation of MBOAT1 and the tetraspanin epithelial membrane protein 2 (EMP2), resulting in the remodeling of the lipidome to favor the incorporation of PUFAs in place of MUFAs. Other investigations have found that cell cycle arrest may generally also enhance resistance to ferroptosis, no matter the trigger, by promoting the sequestration of PUFAs into lipid droplets.¹¹⁴ Separately, p53 expression may also promote ferroptosis independent of its effects on cell cycling by enhancing the activity of the lipoxygenase enzyme ALOX12.^21,115

Cell density and cell-cell contact

Cell density can impact ferroptosis sensitivity. In adherent (2D) epithelial cancer cells and Gpx4-deficient fibroblasts, culturing cells in high cell density can increase resistance to both cystine deprivation and direct GPX4 inhibition.^{30,108,116–118} Mechanistically, this increase in ferroptosis resistance involves the engagement of E-cadherin signaling, activation of the Hippo signaling pathway, reduced levels of iron, ROS, and PUFAs, and increased levels of MUFAs.^116,119,120 On the other hand, despite increased cell density/cell-cell contact, some cells can become more sensitive to ferroptosis when cultured in non-adherent (3D) conditions compared to 2D conditions.¹¹⁶ Furthermore, the propagation of ferroptosis “waves”, a process that appears to occur in the developing chick limb, occurs between neighboring cells that must lie in proximity to one another.⁷ Thus, the effects of cell density and cell-cell contact on ferroptosis sensitivity are context-dependent.

Adhesion to the extracellular matrix

Ferroptosis sensitivity can be altered depending on whether a cell is adhered to substrate. For breast epithelial cancer cells, detachment from the extracellular matrix (ECM) can lead to enhanced ferroptosis resistance. One resistance mechanism involves cell clustering via nectin cell adhesion molecule 4 (NECTIN4, PVRL4) and β4 integrin, which helps maintain GPX4 expression in detached cells.¹²¹ Another resistance mechanism in detached mammary epithelial and breast carcinoma cells involves increased prominin 2 (PROM2)-dependent multivesicular body formation and export of iron from the cell via exosomes.¹²² In breast epithelial tumor organoids, the engagement of β4 integrin by the matrix ligand LM332 promotes ferroptosis resistance.¹²³ When this ligand is absent, these same cells are more sensitive to ferroptosis. Ferroptosis sensitivity in spheroids can also be counteracted by higher NFE2L2/NRF2 expression, which increases intracellular GSH and lowers the accumulation of oxidizing species.¹²⁴ Thus, ferroptosis sensitivity is not only influenced by how densely cells are grown, but also by the adhesion of cells to one another.

Epigenetic cell state

Ferroptosis sensitivity can be altered over time by the dynamic changes in the epigenetic cell state. In general, conditions that promote epithelial-to-mesenchymal transition (EMT) render cancer cells more ferroptosis-sensitive due to enhanced ZEB1-mediated PUFA metabolism.⁸⁹ Drug-tolerant persister cells – rare cells that survive treatment with cytotoxic and targeted anti-cancer agents – also acquire heightened sensitivity to GPX4 inhibition.^90,125 This may be linked to altered redox metabolism or the induction of activating transcription factor 4 (ATF4).⁹¹ A related phenomenon may occur during melanoma dedifferentiation in response to targeted therapies (e.g., BRAF inhibitors) and immunotherapy. Treatment-induced melanoma cell dedifferentiation occurs over several weeks and is associated with epigenetic reprogramming (e.g., loss of MITF expression), lower intracellular GSH levels, and increased sensitivity to ferroptosis-inducing agents.¹²⁶ In triple negative breast cancer (TNBC), the “luminal androgen receptor” subtype contains higher levels of PUFA-containing phospholipids, rendering these cells more dependent on GPX4 for survival compared to other TNBC subtypes.¹²⁷ Other examples of cell state-dependent differences in ferroptosis sensitivity have emerged in the context of small cell lung cancer: the non-neuroendocrine sub-state is more sensitive to ferroptosis compared to the neuroendocrine sub-state due to the former exhibiting a more pro-ferroptotic lipid profile.¹²⁸

Diet

Diet can substantially influence ferroptosis sensitivity. Feeding the nematode worm C. elegans dihomo-gamma-linolenic acid (DGLA) is sufficient to induce ferroptosis in germ cells.^15,83 DGLA toxicity is suppressed by co-feeding with the MUFA oleic acid. DGLA is a type of conjugated PUFA, which is defined by the positioning of their double bonds. Conjugated PUFAs can also act as potent ferroptosis inducers in cancer cells.⁴⁹ Feeding with the conjugated PUFA α-eleostearic acid (αESA) is sufficient to induce ferroptosis in cultured cancer cells and in mouse xenograft tumors.⁵⁰

Manipulation of dietary amino acid levels can modulate ferroptosis sensitivity. In a mouse model of brain cancer, a diet restricted in cysteine and methionine heightened cancer cell sensitivity to GPX4 inhibition compared to animals fed a regular diet.¹²⁹ Dietary effects are likely context-specific: methionine deprivation can also reduce ferroptosis sensitivity when this process is triggered by cystine deprivation.⁷¹ This is because methionine is required to synthesize ChaC glutathione specific gamma-glutamylcyclotransferase 1 (CHAC1), an enzyme that degrades GSH to hasten the onset of ferroptosis.¹³⁰ Engagement of this mechanism could explain why other amino acids like arginine are essential for the execution of ferroptosis in response to cystine deprivation.⁷¹

Dietary antioxidants can also inhibit ferroptosis. When mice are fed a standard laboratory diet, hepatocyte-specific disruption of Gpx4 results in cell death and postnatal lethality.⁴³ However, when mothers are fed a diet enriched in vitamin E, pups are born at normal Mendelian ratios and survive. Thus, differences in diet can impact whether a particular factor is necessary for ferroptosis. This is another example of the contingent nature of ferroptosis sensitivity: a diet enriched in natural antioxidants can create a ferroptosis resistance barrier that is difficult to overcome, even by strong genetic interventions (e.g., Gpx4 deletion).

A DYNAMIC CONTINUUM OF FERROPTOSIS SENSITIVITY

The facts adduced above suggest that ferroptosis sensitivity and resistance can be shaped by many different stable and dynamic factors, including cell lineage, genotype, epigenetic cell state, the metabolic milieu, physical interactions between cells, and interactions between cells and the extracellular matrix. Cells may therefore not have a fixed ferroptosis sensitivity over their lifetime, and different cells in the body, even those from the same lineage or found in the same location, may differ in ferroptosis sensitivity. To help understand how these differences might impact organismal phenotypes, we propose that cells exist on a dynamic ferroptosis sensitivity-resistance continuum (Figure 2), where cells occupy one position on this continuum at one given time, but can shift over time between more resistant and more sensitive states in response to endogenous or exogenous factors. Some changes may radically alter a cell’s position on the continuum, while other changes may have subtler effects. The contingent nature of ferroptosis means that the removal of a single element (e.g., iron) can be sufficient to push an otherwise sensitive cell into a completely resistant state. Other changes, like cell cycle arrest, may increase ferroptosis sensitivity by altering the PUFA to MUFA ratio. Metabolite levels occur in a gradient and can vary considerably between cells of the same type, within the same tissue.¹³¹ Accordingly, we envision that ferroptosis sensitivities will vary in a graded manner. The location of a cell on the ferroptosis sensitivity-resistance continuum will also relate to how ferroptosis is induced. Thus, factors like ACSL4, which seem essential for sensitivity to ferroptosis caused by GPX4 inhibition, may be less important for the positioning of a cell on the continuum when ferroptosis is triggered by cystine deprivation. It might be necessary to envision a range of possible continuum landscapes, depending on the ferroptosis trigger.

Figure 2. The ferroptosis sensitivity continuum.

Cells can have inherent differences in ferroptosis sensitivity stemming from factors such as cell type/lineage, sex, and genotype. However, ferroptosis sensitivity is not entirely fixed by these stable factors. For instance, the extracellular metabolic milieu can provide a cell with molecules that promote or inhibit lipid peroxidation; the proliferative and/or senescent state of a cell can shift the availabilities of pro- or anti-ferroptotic molecules; changes in cell-cell and cell-matrix contacts can trigger signaling pathways that affect iron, oxidizing species, and oxidizable lipid levels; epigenetic cell state shifts can come with gene expression changes that ultimately alter oxidant and antioxidant production; and, diet can modulate the ratios of oxidizable lipids in the membrane, enrich cells with natural antioxidants, alter pro-ferroptotic amino acid levels, or provide the building blocks needed for the synthesis of pro- or anti-ferroptotic proteins. These factors sum up to dictate where a cell may reside within a conceptual continuum of ferroptosis sensitivity.

CONCLUSIONS

The concept of a ferroptosis sensitivity-resistance continuum may help us think about how ferroptotic death events are channeled across a limited range of cells in a developing tissue, why certain cells are more or less susceptible to ferroptosis in disease, and whether we can induce ferroptosis in cancer cells but not in normal cells. In the developing chick limb, a ferroptotic wave is confined to certain muscle cells.⁷ How so? Those cells that transmit the ferroptotic signal and die must exist on the sensitive side of the continuum, while surrounding cells that survive must exist on the resistant side of the continuum. Is it possible that a single factor explains the different positions that these cells take? Or does a combination of factors interact to set the boundary between sensitive and resistant cells? Is the nature of this boundary-setting process specific to the (currently unknown) ferroptosis trigger in these cells, meaning that it may not be generalizable to other developing tissues? All these questions remain to be answered.

Ferroptosis resistance presents a barrier for the treatment of cancer. This barrier may be overcome by diets depleted of anti-ferroptotic molecules like cysteine¹²⁹. New small molecule inhibitors are also being developed for system x_c⁻, FSP1 (ref.^47,48), and GPX4 (ref.^132,133) that can be used in vivo. Inhibitors of anti-ferroptotic lipid metabolic enzymes may augment these effects. It is likely that different cancers will present unique resistance mechanisms that will have to be surmounted to effectively induce cell death. For cancers, the goal of an effective therapy must be to maximize the positioning of cancer cells on the sensitive end of the continuum, while normal cells remain on the more resistant end of the spectrum. At present, only empirical data will allow us to determine how much inhibition of anti-ferroptosis networks can be tolerated in normal versus cancer cells. Ultimately, the construction of a ferroptosis sensitivity map for every cell type in the body, with an understanding of the quantitative contributors of ferroptosis regulation, may be needed to design the most effective therapies.