Molecular mechanisms of ferroptosis and its role in cancer therapy

Abstract Ferroptosis is a newly defined programmed cell death process with the hallmark of the accumulation of iron‐dependent lipid peroxides. The term was first coined in 2012 by the Stockwell Lab, who described a unique type of cell death induced by the small molecules erastin or RSL3. Ferroptosis is distinct from other already established programmed cell death and has unique morphological and bioenergetic features. The physiological role of ferroptosis during development has not been well characterized. However, ferroptosis shows great potentials during the cancer therapy. Great progress has been made in exploring the mechanisms of ferroptosis. In this review, we focus on the molecular mechanisms of ferroptosis, the small molecules functioning in ferroptosis initiation and ferroptosis sensitivity in different cancers. We are also concerned with the new arising questions in this particular research area that remains unanswered.


| INTRODUC TI ON
Ferroptosis is a newly coined non-apoptotic programmed cell death process that was discovered via a chemical screen. 1 The general initiation mechanisms of ferroptosis have partially been elucidated with the research going further. The metabolism of cysteine, polyunsaturated fatty acids (PUFAs) and iron are all closely correlated with ferroptosis initiation. Multiple signalling pathways as well as the cell organelles have also been found to involve in the ferroptosis regulation.
Moreover, a series of small molecules have been found to be able to induce ferroptosis in a wide range of cancer cells. These findings provide the possibility of cancer therapies through genetic or pharmacological interference with ferroptotic cell death, which is of great interest in both scientific research and medicine. Different kinds of cancers seem to have various sensitivities to ferroptosis. A clear understanding of ferroptosis sensitivity in cancers from different tissues will also benefit the clinical practice in applying ferroptosis to cancer therapy. In this review, we summarize the general initiation mechanisms of ferroptosis, the small molecules involved in ferroptosis initiation and the signalling pathways as well as the cell organelles involved in ferroptosis regulation. Moreover, we talk about the potential application of ferroptosis in overcoming cancer cell drug resistance from several aspects.

| Ferroptosis represents a new way of cell death process
Ferroptosis is different from other programmed cell death from several aspects. Lipid peroxides accumulation and iron dependency are the two major features of ferroptosis. Ferroptotic cell death also has unique morphological and bioenergetic features including shrunken mitochondria, increased mitochondrial membrane density, disruption of membrane integrity and depletion of intracellular NADH, but not ATP levels. The induction of ferroptosis depends on ATP production but does not require caspase activation. Moreover, ferroptosis is not sensitive to the inhibition of RIP1/RIP3 or Cyclophilin D, which are key regulators of necrosis, and inhibition of autophagy by 3-MA does not modulate this cell death process. 1 The evidence suggests that ferroptosis is a completely new way of cell death.

| Cysteine metabolism plays a central role in ferroptosis initiation
Erastin is among the small molecules identified in chemical screen that induce ferroptosis in oncogenic RAS mutation cell lines. There were also some other targets of erastin identified in the affinity purification. SLC7A5 was also identified as an erastin binding protein by affinity assays. SLC7A5 can also bind SLC3A2 to form amino acid transporters (system L) of large, neutral amino acids.
However, the inhibition of system L-mediated amino acid uptake by erastin does not contribute directly to ferroptosis. The binding of erastin to the SLC7A5 likely interferes with cystine uptake by the SLC3A2/SLC7A11 complex in trans. 1 Erastin can also target the mitochondrial-resident voltage-dependent anion channel-2 (VDAC2).
Knockdown of VDAC2 or VDAC3 by RNAi attenuates erastin-induced ferroptosis. Erastin-VDAC2 interaction inhibits the permeability of VDAC2 to endogenous substrates, such as NADH and decrease the NADH oxidation in cancer cells, which induces mitochondrial dysfunction and the release of oxidative species (Figure 1).
However, both VDAC2 and VDAC3 are necessary, but not sufficient, for erastin-induced cell death showed by the knockdown or overexpression studies. 7

| GPX4 inactivation is causative for lipid peroxides accumulation
RSL3 is another molecule for ferroptosis initiation found in the chemical screen. 1 The target for RSL3 was investigated by proteomic analysis of an affinity pull-down assay, and GPX4 was found to be a direct target of RSL3. 8 GPX4 can prevent the toxicity of lipid peroxides by its enzyme activity and maintain the homeostasis of membrane lipid bilayers ( Figure 1). 9 RSL3 inhibits the activity of GPX4 by covalent bonding with GPX4 and leads to lipid peroxides accumulation ( Figure 1). Ferroptosis induced by RSL3 treatment is similar to that of GPX4 inactivation, further supporting that RSL3 induces ferroptosis through GPX4 inhibition. GSH is the co-factor of GPX4 in catalysing peroxides into alcohols. 10 GSH depletion caused by cysteine deprivation directly inactivates GPX4 and leads to subsequent induction of ferroptosis ( Figure 1).

| Glutaminolysis is indispensable for cysteine deprivation-induced ferroptosis initiation
Recent studies have shown that glutaminolysis played an indispensable role in ferroptosis initiation in mouse embryonic fibroblasts (MEFs). 17 Glutamine is degraded via glutaminolysis and the tricarboxylic acid (TCA) cycle. 17,18 Further evidence demonstrates that glutaminolysis metabolite αKG or its downstream metabolites during the TCA cycle are required for the induction of ferroptosis ( Figure 1). 17 Several enzymes involved in glutaminolysis have been revealed to play important roles in ferroptosis initiation. Transaminase converts glutamate into αKG through the transamination process. 19 The inhibitor of transaminases, amino-oxyacetate (AOA), was found to inhibit ferroptosis in MEFs. Knockdown of the transaminase GOT1 could also inhibit cysteine deprivation-induced ferroptosis in MEFs ( Figure 1). 17,20 However, the role of glutaminolysis in ferroptosis regulation is more complex. Glutamate dehydrogenase 1 (GLUD1) converts glutamate into αKG through the glutamate deamination.
However, RNAi-mediated knockdown of GLUD1 failed to inhibit ferroptosis initiation. 17 Both Glutaminases 1 (GLS1) and Glutaminases 2 (GLS2) catalyse the conversion of glutamine into glutamate. Studies demonstrated that only GLS2 is involved in the regulation of ferroptosis ( Figure 1). Further study illustrated that GLS2 is the transcriptional target of p53 and is up-regulated during p53-dependent ferroptosis. 20,21 Moreover, only in combination with glutamine can cysteine deprivation induce ROS accumulation, lipid peroxidation and ferroptosis. This phenomenon was explained by recent study F I G U R E 1 Mechanisms of ferroptosis induction. Inhibition of system x c − deprives cellular cysteine, leading to GSH deletion and GPX4 inactivation. GSH can be synthesized from methionine through the transsulphuration pathway which is inhibited by cysteinyl-tRNA synthetase. RSL3 inhibits the activity of GPX4 by covalent binding with GPX4. GPX4 inactivation leads to the accumulation of lipid peroxides and final ferroptosis. Enzymes (GLS2 and GOT1) involved in glutaminolysis regulate ferroptosis process. The tricarboxylic acid (TCA) cycle promotes cellular GSH deletion and leads to ferroptosis in combination with cysteine deprivation. The mitochondrial genes (ACSF2, CS) are all involved in ferroptosis regulation. ER stress induced by ferroptotic reagents promotes ferroptosis through ATF4-dependent CHAC 1 expression. Lysosome is also involved in ferroptosis induction through autophagy process or cathepsin B release. Lysosome ROS contributes to the lipid ROS production from Gao et al, 22 whose work discovered a role for mitochondria in cysteine deprivation-induced ferroptosis in both MEFs and HT1080 cells. Under cysteine deprivation, glutaminolysis will promote mitochondrial respiration and the rapid exhaustion of GSH by GPX4, inducing potent ferroptosis. However, when glutaminolysis is inhibited, GSH turnover rate is slowed and ferroptosis is not induced even when cysteine is deprived. (Figure 1). 22 Elevated glutaminolysis has been observed in most cancer cells to satisfy their bioenergetic requirements. However, high rate of glutaminolysis showed vulnerability of cancer cells for its role in promoting ferroptosis induction in cancer cells. Glutaminolysis in combination with cysteine deprivation induces potent ferroptotic cell death and will revolutionize the anti-tumour strategy.

| PUFAs and cellular iron are essential for lipid peroxide accumulation
Polyunsaturated fatty acids increase membrane fluidity and are important for the adaption of original life to the environment.
However, PUFAs can be oxidized by intracellular ROS and produce the lipid peroxides that promote the induction of ferroptosis ( Figures   1 and 2). 23 The activity of lipoxygenases (LOXs) could catalyse PUFA-containing phospholipids into pro-ferroptotic lipid peroxidation. 24 A CRISPR-based genetic screen identified two lipid metabolism regulators, lysophosphatidylcholine acyltransferase 3 (LPCAT3) and acyl-CoA synthetase long-chain family member 4 (ACSL4) that promote GPX4 inhibition-induced ferroptosis in KBM7 cells. 25 The catalysed functions of ACSL4 and LPCAT3 are responsible for membrane phospholipid insertion and polyunsaturated fatty acid remodelling ( Figure 2). 26,27 Knockout of ACSL4 resulted in marked resistance to ferroptosis induced by GPX4 deficiency. 28 Iron serves as the essential component of many enzymes involved in DNA synthesis, metastasis, cell circle progression or angiogenesis ( Figure 2). However, iron is also a redox-active reagent and promotes ROS production via Fenton reaction ( Figure 2). 29  stabilizes it and increases its expression or changes its localization ( Figure 2). 31 Transferrin is the transporter of iron, and it imports iron into the cell from the extracellular environment through recognition by TFR1 ( Figure 2). Both transferrin and TFR1 are the targets of IRP1 and IRP2 and are required for ferroptosis induction. 32 Ferritin is also the target of IRP1 and IRP2. Ferritin binds to the free iron and makes it unavailable and therefore functions to prevent ferroptosis. 33,34 Ferroportin is responsible for iron export from the cells and is the negative regulator of ferroptosis ( Figure 2). 35 Iron chelators have long been applied to the development of anti-tumour strategies. 36 However, the iron requirement in ferroptosis induction redefined the role of iron chelators in cancer treatment. Future studies are still needed to validate the iron chelation strategies in cancer therapy under different conditions.

| The emerging roles of different organelles involved in ferroptosis regulation
The organelles are the important components of the cell and function to maintain intracellular homeostasis. However, dysfunction of the cell organelles under stress conditions will promote the cell death process. The mitochondria, lysosome and the endoplasmic reticulum (ER) have all been demonstrated to play important roles in ferroptosis regulation. Here, we talk about the specific role of different organelles in ferroptosis regulation.

| The role of mitochondria in ferroptosis
Mitochondria is the energy provider of the cell and has long been considered to be closely related to the programmed cell death process. 37 (Figure 1). 41 A recent study recognized ferroptosis as an autophagic cell death which further reveals the important roles of lysosome in ferroptosis as lysosome is the major organelle for autophagic degradation of protein aggregates. [42][43][44] Inhibition of autophagy by ATG13 and ATG3 knockdown greatly reduced cysteine deprivation-induced ferroptosis. 39 Autophagy may function to promote the ROS production and subsequent accumulation of lipid peroxides. Degradation of ferritin through NCOA4-mediated ferritinophagy will release the iron for ferroptosis induction ( Figure 1). NCOA4 knockdown decreases the ferritinophagy and leads to the unavailability of free iron, which abrogates the accumulation of ROS and decreases the ferroptosis induction. 45 However, there is still lack of evidence to demonstrate ferroptosis as a direct consequence of autophagy.

| ER in ferroptosis
Endoplasmic reticulum stress is induced under various pathological conditions and is closely related to the cell death process. The ER stress responsive genes promote the apoptosis or autophagy process. 46,47 Erastin can also induce ER stress and up-regulates ER stress responsive genes. 5,48,49 The eif2α-ATF4 branch is the major signalling pathway activated by ferroptotic reagents (Figure 1). CHAC 1 is the downstream of ATF4 and demonstrated to promote the degradation of GSH and the subsequent ferroptosis ( Figure 1). 5,50 PUMA is another downstream of ATF4 and is also up-regulated during the ER stress induced by the ferroptotic reagents, artemisinins (ART). However, PUMA activation will induce apoptotic cell death under the treatment of ferroptotic agents. What's more, knockout of PUMA did not reduce cell death by ART treatment alone. Only in combination treatment of ART and an apoptosis-inducing ligand, TRAIL, did knockout of PUMA reduce cell death. So, the role of PUMA in ferroptosis is still elusive (Figure 1). Evidence showed that the ER stress induced by the ferroptotic reagents cannot be relieved through the inhibition of lipid peroxides by the Fer-1 and Lip-1. 49 There is still little evidence to demonstrate the direct role of ER stress responsive genes in ferroptosis regulation.

| Signalling pathways in ferroptosis regulation
In addition to those key ferroptotic initiation signals, multiple pathways are also involved in ferroptosis regulation. We summarize these ferroptotic signalling pathways associated with cancer progression and reveal its potential application in cancer therapy (Table 1). We also talked about several important signalling pathways in the following text.

| Nrf2
Nrf2 is a transcription factor that regulates iron metabolism genes

| P53
P53 is a tumour suppressor gene that is activated under different stress stimuli. P53 is involved in ferroptosis as a transcriptional repressor of SLC7A11, impairing cysteine import and promoting ferroptosis initiation. [54][55][56] P53 is also involved in other programmed cell death processes. However, the mechanism of p53 in ferroptosis induction is specific and different from other already known programmed cell death. An acetylation-defective p53 mutant, p53 3KR , has been created with 3 lysine residues replaced by arginine residues. 57 This mutant is highly effective in repressing SLC711A expression but not that of other already known p53 target genes (cell cycle, apoptosis or senescence-related genes).
However, the acetylation-defective form of p53, p53 4KR98 is unable to inhibit SLC711A expression, while this mutated form can still repress the other p53 target genes. 54

| Haeme oxygenase-1
Haeme oxygenase-1 can be regulated both by the transcriptional factor Nrf2 and the endoplasmic reticulum-associated degradation pathway (ERAD). 59 These results suggest a dual role of HO-1 in ferroptosis induction.

| FANCD2
Ferroptosis is involved in bone marrow injury caused by the traditional cancer therapy. FANCD2 is a nuclear protein involved in DNA damage repair, and its role in ferroptosis induction during the bone marrow injury was recently validated. 65

Reagents Target Mechanisms References
Erastin and its analogs

| Small molecule inducers of ferroptosis
Ferroptosis was originally defined during a chemical screen for cancer treatment. With increased research on ferroptosis, more ferroptosis-inducing compounds have been identified. We summarize the existed compounds in ferroptosis induction in Table 2 and its applications in different cancer cells in Table 3.

| Ferroptosis regulation in different cancers
Although the precise mechanism that determines the ferroptosis sensitivity in cancer cells is largely unknown, cancer cells from different tissues show different degrees of ferroptosis sensitivity. We discussed the sensitivity of ferroptosis to the ferroptotic reagents and the general initiation mechanism of ferroptosis above, and next, we talk about the specific mechanisms of ferroptosis induction in several special types of cancer cells.

| Ferroptosis in breast cancer cells
Breast cancer cells seem less sensitive to the ferroptotic reagents

| Ferroptosis in lymphomas and renal cancers
In a study performed by Wan Seok Yang, 8

| Ferroptosis in brain tumours
The nervous system contains the highest content of PUFAs in our body, which are the main substrates for the production of peroxides. Brain tumours were more sensitive to ferroptosis induction. Both erastin and sorafenib can induce potent cell death in malignant brain tumours. However, brain tissues also develop a protection system against cell death. Increased Nrf2 activation was also observed in brain tumours. The Nrf2-Keap1 pathway protects cancer cells from ferroptosis induction as mentioned above. In glioma cells, Nrf2 overexpression or Keap1 knockdown promotes the oncogenic transformation. Nrf2 up-regulation in patients with brain tumours also showed reduced survival rate.

| The potential application of ferroptosis in overcoming cancer cells' drug resistance
Cancer cells' resistance to chemotherapy is a major problem in cancer treatment. The ineffective induction of cell death is one of the features shared by most chemotherapy drugs. As ferroptosis is a totally different cell death process from apoptosis, ferroptotic reagents may represent a promising strategy in overcoming the inefficiency of apoptosis-inducing chemotherapy drugs in cell death induction. Efforts have been made to explore the application of inducing ferroptosis in overcoming the cancer cells' drug resistance.

| Ferroptosis promotes the cell death of drugtolerant cancer cells with mesenchymal state
Transition of epithelial cancer cells into a mesenchymal state via epithelial to mesenchymal transition (EMT) is a process that brings multiple mechanisms of resistance to cell death including the inactivation of apoptotic programmes across a large range of cancer cells. 78,79 Insight into vulnerabilities of these cancer cells with mesenchymal state is a promising way to improve the therapeutic strategy. Studies showed that cancer cells with mesenchymal state harboured a higher activity of enzymes that promote the synthesis, storage and use of long-chain PUFAs, which are the sources of reactive lipid peroxides, making these cancer cells highly dependent on GPX4 for survival ( Figure 3). 80 This vulnerability makes it possible to induce ferroptosis in these cancer cells through inhibition of GPX4. In fact, the ferroptotic reagents are demonstrated to strongly correlate with the selective cell death of epithelial-derived cancer cells with high-mesenchymal state (Figure 3). 80  These results can be verified in different types of cancer cells.
Evidence has shown that HCC4006 non-small cell lung cancer cells with a high-mesenchymal state are resistant to gefitinib. However, these same cells were preferentially sensitive to GPX4 inhibition compared with parental cells. Cancer cells with mesenchymal origin also showed great sensitivity to ferroptotic compounds. 80

| Ferroptosis promotes the cell death of the drug-tolerant persister cancer cells
The potential application of ferroptosis in overcoming cancer cells' drug resistance can also be reflected from its role of inducing cell death of persister cells. Persister cells are the surviving cancer cells upon treatment with several rounds of chemotherapy drug, which is another therapy-resistant cell state presented across a wide range of tumour types ( Figure 3). 82,83 Targeting the persister cancer cells is also an important strategy for overcoming cancer cells' drug resistance. The stemness markers and mesenchymal markers are upregulated in the persister cells, showing the mesenchymal state of these cancer cells. 84 The exploration of the vulnerability of these cancer cells shows that the Nrf2 target genes were down-regulated.
As we mentioned above, Nrf2 is a major suppressor of ferroptosis.
Further study showed that persister cells have markedly decreased levels of both glutathione and NADPH and have a specific sensitivity to lipid peroxidation rather than general sensitivity to oxidative stress. Evidence shows that GPX4 inhibitors are specifically lethal in the persister cells through ferroptotic cell death ( Figure 3). 80,84 Based on these results, inducing ferroptosis may be a promising way to overcome these cells' drug resistance. Nrf2 target genes are down-regulated, and the levels of NADPH and GSH are decreased in these cells with mesenchymal state. GPX4 inactivation is lethal to cancer cells with mesenchymal state explain the role of iron in promoting ROS production. However, this reaction is not specific to the iron. Other metal irons also contain this feature but are not responsible for the ferroptotic induction.

| DISCUSS IONS
Ferroptosis-inducing compounds are exclusively effective to certain cancer cells but not others. Efforts are still needed to classify the types of cancers which are sensitive to ferroptosis. This is important for the application of ferroptosis to cancer therapy. Moreover, the relationship between ferroptosis and other cell death process needs to be classified clearly under different pathological conditions because it is important to combine different methods to the diseases treatment.

ACK N OWLED G EM ENTS
The authors thank all laboratory members for ongoing discussions.

CO N FLI C T O F I NTE R E S T
The authors confirm that there are no conflicts of interest.

AUTH O R CO NTR I B UTI O N
TX and JW provided direction and guidance throughout the prepara-