Ferroptosis: Biochemistry and Biology in Cancers

The challenge of eradicating cancer is that cancer cells possess diverse mechanisms to protect themselves from clinical strategies. Recently, ferroptosis has been shown to exhibit appreciable anti-tumor activity that could be harnessed for cancer therapy in the future. Ferroptosis is an iron-dependent form of regulated cell death that is characterized by the oxidization of polyunsaturated fatty acids (PUFAs) and accumulation of lipid peroxides. Ferroptosis has been closely correlated with numerous biological processes, such as amino acid metabolism, glutathione metabolism, iron metabolism, and lipid metabolism, as well as key regulators including GPX4, FSP1, NRF2, and p53. Although ferroptosis could be involved in killing various cancer cells, multiple aspects of this phenomenon remain unresolved. In this review, we summarize the biochemistry and biology of ferroptosis in diverse cancers and discuss the potential mechanisms of ferroptosis, which might pave the way for guiding cancer therapeutics.


INTRODUCTION
Cancer has been widely recognized as the second leading cause of death worldwide (1). According to recent statistics on cancers in 185 countries, there were 19.3 million new cancer cases (18.1 million excluding nonmelanoma skin cancer) and 10 million cancer deaths (9.9 million excluding nonmelanoma skin cancer) worldwide in 2020 (2). Although several therapeutic strategies such as surgery, radiotherapy, chemotherapy, ablation, and organ transplantation have been utilized to fight against cancer, many malignant cancers still do not show satisfactory results after treatment. The two major causes of cancer-related death are metastasis and recurrence of tumors; therefore, pinpointing mechanistic details of the pathways by which cancer cell death occurs is crucial for identifying treatment strategies and guiding clinical therapy.

OVERVIEW OF IDENTIFICATION OF FERROPTOSIS
In 2012, the concept of ferroptosis was first described by Scott J. Dixon (a member of the Brent R. Stockwell group) and his collaborators who described the characteristics of ferroptosis (3). Ferroptosis, a novel type of regulated cell death (RCD), is a unique form of intracellular iron-dependent peroxidation of PUFA-containing phospholipids (PLs), and is morphologically, biochemically, and genetically distinct from other forms of RCD including apoptosis, autophagy, and necroptosis. Cells undergoing ferroptosis show unique hallmarks including rupture of cellular membranes, smaller mitochondria with increased mitochondrial membrane density, reduced/vanished mitochondria cristae, rupture of outer mitochondrial membranes, and a normal nucleus (29).

Early Studies Related to Ferroptosis
In 1955, Eagle et al. first found that human uterine carcinoma HeLa cells cultured without cystine exhibited a unique microscopic morphology that was different than that resulting from deprivation of other amino acids (30). They also found that cells cultured in cystine-free medium failed to grow but could be restored by supplementing them with glutathione (GSH) (31,32). In 1977, Bannai et al. showed that cystine starvation of human lung fibroblasts resulted in rapid reduction of GSH and subsequent cell death; however, cell death could be rescued by the addition of the lipophilic antioxidant a-tocopherol (a component of vitamin E) (33). These results implied that cystine could sustain the intracellular level of GSH and that there might be an accumulation of reactive oxygen species (ROS) that could be prevented by lipophilic antioxidants.
In 1965, two separate research teams both identified lipid peroxidation as a prime cause of cellular damage in rat liver (34,35). In the 1980s, lipid peroxidation was considered to be one of the main forms of oxidative damage via the destruction of unsaturated lipid components of cell membranes and lipoproteins in some pathologies (36,37). Nonetheless, these discoveries were considered as mechanisms of cellular damage at that time.

Conceptualization of Ferroptosis
Brent R. Stockwell and members tried to screen small molecules that could selectively kill cells overexpressing the oncogenic mutant HRAS. In 2003, they identified a novel compound that they named "erastin", and explored the effect of erastin in engineering tumor cells. However, they found that no characteristics of apoptosis occurred, such as caspase activation, cleavage of caspase substrates, annexin V staining, and morphological changes in the nucleus (38). In 2007, they further reported that erastin induced the formation of oxidative species and subsequent death through an oxidative nonapoptotic mechanism, and that the cell death induced by erastin could be suppressed by a-tocopherol (39). In 2008, they reported yet another small compound, Ras selective lethal 3 (RSL3), which induced a similar iron-dependent non-apoptotic cell death in oncogenic RAS-harboring cancer cells, which could also be suppressed by both a-tocopherol and desferrioxamine mesylate (DFOM) (40). In 2011, the authors distinguished erastin-and RSL3-induced cell death from the mechanism of action of other cell death inducers (41). In 2012, they named this phenomenon of erastin-induced iron-dependent cell death as ferroptosis (3).

MECHANISMS OF FERROPTOSIS
As an increasing number of groups actively investigate ferroptosis, numerous mechanisms have been found to be involved in ferroptosis, such as amino acid metabolism, GSH metabolism, iron metabolism, and lipid metabolism ( Figure 1 and Table 1).

Amino Acid and GSH Metabolism
GSH functions as a cofactor of GSH peroxidase 4 (GPX4, discussed in detail below) in lipid metabolism; thus, the synthesis of GSH regulates ferroptosis. GSH is synthesized with three substrates, namely glutamate, cysteine, and glycine, in two steps. The initial and limiting step is the combination of glutamate and cysteine to form GluCys in the presence of glutamate-cysteine ligase (GCL, previously known as gglutamyl-cysteine synthetase) utilizing ATP. Then, GluCys combines with glycine to yield GSH, which is catalyzed by GSH synthetase (GSS) in the presence of ATP. Buthionine sulfoximine (BSO) is a small molecular inhibitor of GCL known to indirectly inhibit the enzymatic activity of GPX4 (42). The pentose phosphate pathway (PPP) generates NADPH, which is essential for preserving the cellular levels of GSH. Inhibition of the PPP and silencing of two PPP enzymes, g l u c o s e -6 -p h o s p h a t e d e h y d r o g e n a s e ( G 6 P D ) a n d phosphoglycerate dehydrogenase (PGD), hinders erastininduced ferroptosis in human lung cancer cells (3).
by preventing cystine import (3). Cysteine could also be imported into cells through the SLC7A10 transporter (66). Some cells utilize the trans-sulphuration pathway to biosynthesize cysteine from methionine through methionine-R-sulfide reductase B2 (MSRB2), that bypasses the requirement for cystine import via system X − c (66,67). Pharmacological and genetic inhibition of SLC7A11 induces ferroptosis and significantly sensitizes cisplatin-resistant head and neck cancer (HNC) cells to chemotherapy in vitro and in vivo (68). Deletion of SLC7A11 induces tumor-selective ferroptosis and inhibits pancreatic ductal adenocarcinoma (PDAC) growth (43).
Glutamine, transported into cells through SLC1A5, is also essential for the induction of ferroptotic cell death. Although glutaminase 1 (GLS1) and GLS2 are structurally and enzymatically similar and can catalyze the conversion of glutamine into glutamate, only GLS2 is required for ferroptosis (46). Upregulation of GLS2, a transcriptional target of the tumor suppressor p53, could induce p53-dependent ferroptosis (44). The glutamic-oxaloacetic transaminase 1 (GOT1) synthesizes aketoglutarate from glutamate, and glutaminolysis is supposed to modulate ferroptosis in the mitochondrial tricarboxylic acid (TCA) cycle (45).

Iron Metabolism
The first observation that iron participates in the process of ferroptosis originated from reports that iron chelators could prevent cell death induced by cystine deprivation (considered as ferroptosis now) in 1996 (69). The biological function of iron is mainly dependent on its ability to accept and donate electrons while switching among the ferrous bivalent (Fe 2+ ), ferric trivalent (Fe 3+ ), and ferryl tetravalent (Fe 4+ ) states by catalyzing various biochemical reactions (70 (72).
Sufficient levels of free intracellular iron is a necessity for triggering ferroptosis. There are four following routes for cells to import iron (73). The most crucial route is mediated by transferrin and its partner, the transferrin receptor 1 (TFR1). Transferrin, a type of globulin mainly synthesized by the liver and released into the serum, has an excellent ability to chelate two Fe 3+ ions. When the transferrin reaches the cell membrane, it identifies and binds to TFR1 and the transferrin-TFR1 complex is internalized via clathrin-mediated endocytosis, whereas apo-transferrin (iron-free transferrin) is not identified by TFR1 and internalized. Fe 3+ is liberated from the transferrin-TFR1 complex at the low pH conditions of endocytic vesicles and reduced to Fe 2+ by the six-transmembrane epithelial antigen of prostate (STEAP) family (73,74). Fe 2+ is then released into the cytoplasm via the divalent metal transporter 1 (DMT1) and participates in establishing the intracellular labile iron pool (LIP) (75). Meanwhile, apo-transferrin and TFR1 remain bound until the complex is recycled back to the plasma membrane, followed by apo-transferrin releases (76). The second route is directly assimilating the free iron unbound by transferrin; however, each iron transporter only transports Fe 2+ . Therefore, Fe 3+ is first reduced to Fe 2+ by several types of ferrireductases (e.g., STEAP, cytochrome B reductase 1, and ferric chelate reductase 1), and the Fe 2+ is directly transported into cells by cellular membrane transporters including DMT1, Zrt-and Irt-like protein 8 (ZIP8), and ZIP14 (73). The third pathway involves the uptake of hemoglobin-containing porphyrin-bound Fe 2+ , especially in macrophages (77). The fourth mechanism involves the assimilation of the iron-storage protein ferritin by its cellular membrane receptors such as the scavenger receptor class A member 5 (SCARA5) found in embryos and in the kidney (78). After iron uptake, the intracellular iron can be utilized, stored, and exported. In terms of utilization, mitoferrin 1 and 2 specially transport iron into the mitochondria to assist with cellular respiration and synthesis of Fe-S clusters and heme (79). There are two forms of intracellular storage, LIP and ferritin. LIP causes active oxidative stress-related toxicity and is responsible for regulating cellular iron homeostasis through the iron regulatory protein-iron responsive element system. Ferritin is an iron-sequestering protein that contains up to 4500 iron atoms and possesses multiple functions in iron delivery, cell proliferation, angiogenesis, and immunosuppression (80). Ferritin can also be degraded to release free iron via the nuclear receptor coactivator 4 (NCOA4), a process termed as "ferritinophagy" (81). In terms of iron export, ferroportin (FPN, i.e., SLC40A1), once known as the only iron efflux pump that cooperates with ceruloplasmin or hephaestin, is mainly responsible for transporting Fe 2 + out of cells (82). Ceruloplasmin suppresses ferroptosis by regulating iron homeostasis in HepG2 and Hep3B cells, and depletion of ceruloplasmin results in the accumulation of intracellular Fe 2+ and lipid ROS and promotes erastin-and RSL3-induced ferroptotic cell death (83). In addition, prominin 2 promotes the formation of ferritin-containing multivesicular bodies and exosomes that transport iron out of the cell to facilitate ferroptosis resistance in breast cancer (84).
Treating cells without transferrin or with TFR RNAi could not lead to significant ferroptotic cell death, indicating that an external iron source is required for ferroptosis (46). STEAP1 and STEAP2 are highly expressed in various human cancer types, such as the colon, breast, cervix, prostate, pancreas, bladder, ovary, testis, and Ewing sarcoma (85)(86)(87). STEAP3 is overexpressed in malignant gliomas and induces cancer epithelial-mesenchymal transition (EMT) (88). STEAP4 is activated under hypoxic conditions, leading to mitochondrial iron imbalance and enhanced ROS production (89). FPN is dramatically suppressed in many cancer types (90), indicating that there might be abundant iron in cancer cells. Reduced FPN levels in triple-negative breast cancer (TNBC) cells stimulate proliferation and EMT (91). Hepcidin, synthesized by tumors or the liver, facilitates FPN degradation and contributes to cancer proliferation and progression (92). Proteins involved in raising the levels of intracellular iron (TFR1, DMT1, and hepcidin) are extensively upregulated in tumor cells, whereas proteins suppressing intracellular iron accumulation (FPN and hephaestin) are downregulated (93).

Lipid Metabolism
The fact that ferroptosis is driven by peroxidation of PUFAs, and not monounsaturated fatty acids (MUFAs) or deuterated PUFAs, was confirmed by Brent R. Stockwell and coworkers in 2016 (4).
As is known to all, PUFAs, as well as essential fatty acids, could not be physiologically synthesized and must be obtained from food. PUFA is defined as a fatty acid containing at least two -CH=CH-groups, such as linoleic acid, linolenic acid, and arachidonic acid (AA). PLs, the basic components of the cellular membrane, are classified as glycerophospholipids and sphingomyelin. The predominant fatty acid moiety in sphingomyelin is saturated fatty acids or MUFAs, whereas glycerophospholipids contain an esterified C 16 or C 18 saturated fatty acid at C-1, an esterified C 18 to C 20 unsaturated fatty acid (e.g., AA) at C-2, and a phosphatidic ester group at C-3 of the glycerol backbone. The incorporation of de novo synthesized fatty acids into PLs requires acyl-CoA synthetases (ACSs) that convert long-chain fatty acids to acyl-CoA and lysophospholipid acyltransferases (LPLATs) that, in turn, catalyze the subsequent reacylation step to form PLs (94). For instance, lipidomic studies suggest that phosphatidylethanolamines containing AA or adrenic acid (AdA), are key PLs that undergo oxidation and drive the occurrence of ferroptosis (50,51). ACS long-chain family member 4 (ACSL4) preferentially converts AA to acylated AA, and lysophosphatidylcholine acyltransferase 3 (LPCAT3) subsequently catalyzes the incorporation of acylated AA into the PL (51,95).
There are two mechanisms of lipid peroxidation, the nonenzymatic free radical chain reaction and the enzymatic process. Compared with the tightly controlled enzymatic lipid peroxidation, the free radical chain reaction involving Fenton chemistry is poorly controlled. There are three types of lipid oxidation enzymes, namely cyclooxygenase (COX), cytochrome P450 (CYP), and lipoxygenase (LOX). COXs synthesize lipid endoperoxides and are partially responsible for the peroxidation of linoleic acid, while CYPs synthesize epoxyeicosatrienoic acids and LOXs contribute predominantly to the synthesis of lipid hydroperoxides.
Nonenzymatic lipid peroxidation involves the oxidation of PUFAs, initiated by ROS. In unsaturated acyl chains, the allylic hydrogen atoms, i.e., the hydrogen atoms on methylene groups adjacent to double bonds, exhibit low carbon-hydrogen (C-H) bond energies, while the hydrogen atoms located on methylene between two double bonds (bis-allylic hydrogen atoms) have even lower C-H bond energies, thus bis-allylic hydrogen atoms can be abstracted by ROS, forming PL radical species (PL • ) with the radical centered on the allylic carbon atom (96). Therefore, PUFAs are highly susceptible to oxidative damage due to the existence of bis-allylic hydrogen atoms that saturated fatty acids and MUFAs do not possess. According to previous reports (96,97), in the presence of HO • derived from the Fenton reaction, the PUFA of PL donates a hydrogen atom to HO • and becomes a carbon-centered PL • that further reacts with intracellular molecular O 2 to form a PL peroxyl radical (PLOO • ). Next, with the participation of PUFA moiety from another PL, PLOO • abstracts a hydrogen atom and subsequently converts to PL hydroperoxide (PLOOH) accompanied with a new PL • . PLOOH is cleaved in the presence of Fe 2+ to form the PL alkoxyl radical (PLO • ) that reacts with PUFA of another PL to form PL alcohol (PLOH) and a new PL • , followed by another lipid radical chain reaction.
Lipid peroxidation of membranes substantially alters the physical properties of lipid bilayers in terms of disrupted ion gradients, decreased membrane fluidity, slower lateral diffusion, and increased membrane permeability (106). There exist two mechanisms for eliminating lipid peroxides from cells without the generation of new radicals. GPXs, especially GPX4, are crucial regulators for reducing lipid peroxides to corresponding alcohols by utilizing GSH as a cofactor (107), thereby limiting the transition metal-dependent formation of toxic radicals (e.g., PLO • ). Another mode of degradation involves the conversion of lipid peroxides to aldehydes such as malondialdehyde (MDA) and 4-hydroxynonenal (4-HNE), which should undergo multiple oxidization steps. However, there is still no definitive consensus about the metabolic route that predominantly contributes to the degradation of lipid peroxides among multiple biosynthesis routes (101). MDA could react with primary amines on proteins or DNA to form crosslinks. Moreover, excessive MDA generation within cells is associated with major human diseases including cancer (108). Decomposition of AAs and longer PUFAs generates 4-HNE, which contains three functional groups: (i) an electrophilic C=C double bond that is a Michael acceptor and that forms covalent adducts with nucleophilic amino acids; (ii) an aldehyde that can form Schiff base adducts with primary amines; and (iii) a hydroxyl group that can be oxidized to an electrophilic ketone (109). The electrophile 4-HNE has been widely studied as a signaling molecule that stimulates the cell cycle and cellular proliferation, along with investigation into its cytotoxic nature known to inhibit gene expression (110).

GPX4
According to phylogeny, the GPX family is divided into three groups: GPX1 and GPX2; GPX3, GPX5, and GPX6; GPX4, GPX7, and GPX8, and could also be classified as selenocysteine-containing GPXs and cysteine-containing GPXs depending on whether the active center contains Sec or Cys (111). GPX4, i.e., PL hydroperoxide GPX, is the second selenoperoxidase to be isolated and is regarded as a peroxidation-inhibiting protein after the identification of GPX1 (112). GPX4 is the only GPX for which PL hydroperoxides in membranes and protein-thiol groups can act as oxidizing and reducing substrates, respectively, under conditions of GSH deprivation (113). It was first isolated and purified from pig liver in 1982, and was identified to confer protection to liposomes and biomembranes from iron-catalyzed lipid peroxidation (114). The human GPX4 was isolated from human liver in 1994 (115). GPX4, containing 170 amino acids, has a typical thioredoxin motif consisting of four a-helices that are localized at the protein surface and seven b-strands, five of which form a central b-sheet (116). The catalytic triad, consisting of selenocysteine (U46), glutamine (Q81), and tryptophan (W136) residues, is localized at the protein surface, and mutations in any of these residues could lead to the inactivation of GPX4 (117). For instance, replacement of selenocysteine with cysteine could diminish the activity of GPX4 by 90% and strengthen its sensitivity to redox stress (118). GPX4 reduces lipid peroxides to the corresponding alcohols via oxidation of its active site selenol (Se-H) to selenenic acid (Se-OH), which is then reduced by two equivalents of GSH to the active Se-H (101).
RSL3 has four diastereomers, but only 1S,3R-RSL3 can induce ferroptosis by suppressing the enzymatic activity of GPX4 via covalent modifications without affecting intracellular GSH levels (42,119). Conversely, overexpression of GPX4 could prevent RSL3-induced ferroptotic cell death (42). RSL3 induces ferroptosis in colorectal cancer cells due to decreased expression of GPX4 and both increased levels of ROS and cellular LIP (52). In addition to RSL3, ML162, ML210, and altretamine (an FDA-approved anticancer agent) can also induce ferroptosis by suppressing GPX4 (120). Inducible GPX4 inactivation could also lead to 12/15-LOX-derived lipid peroxidation as a specific downstream event, which indicates that GPX4 could prevent cell death by inhibiting 12/15-LOX (113). FIN56, derived from CIL56, exhibits greater potency and better oncogenic RAS selectivity than CIL56 (121). FIN56 induces ferroptosis by degrading GPX4 protein, instead of downregulating GPX4 mRNA transcription and subsequent synthesis of GPX4 protein, and simultaneously causes depletion of mevalonate-derived coenzyme Q 10 (CoQ 10 ), which is an electron carrier in the mitochondrial respiratory chain and an endogenous antioxidant. FINO2, an endoperoxide-containing 1,2-dioxolane, can initiate ferroptosis selectively in engineered cancer cells, such as renal cancer cell Caki-1 and fibrosarcoma cell HT-1080 (122). FINO2 does not inhibit the antiporter system X − c ; instead, it directly represses the enzymatic activity of GPX4, or depletes GPX4 and CoQ 10 . However, it can directly oxidize iron and indirectly suppresses the enzymatic function of GPX4, ultimately causing widespread lipid peroxidation. The side effects of statin treatment could be attributed to reduced levels of GPX4 in some tissues (123), while the decreased incidence of several cancers on statin consumption (124) and the loss of GPX4 detected in some cancer cells with statin treatment could be ascribed to repression of GPX4 synthesis (125). Further, statin treatment has also been shown to deplete CoQ 10 (126).

FSP1
Ferroptosis suppressor protein 1 (FSP1), previously known as apoptosis-inducing factor mitochondrial-associated 2 (AIFM2), was recently identified as a ferroptosis resistance factor (54). Although GPX4 is essential for cancer cells to escape from ferroptosis, inhibition of GPX4 fails to trigger ferroptosis in some cancer cells regardless of ACSL4 expression, suggesting the existence of additional mechanisms of ferroptosis resistance. The antiferroptotic function of FSP1 is independent of cellular GSH level, GPX4 activity, ACSL4 expression, and oxidizable fatty acid content, showing that FSP1 does not interfere with canonical ferroptosis mechanisms. Moreover, p53 status does not affect FSP1 expression. FSP1 expression is positively correlated with ferroptosis resistance in many cancer cell lines (127). FSP1 can prevent PL peroxidation and its suppression has been shown to be enhanced in the presence of both CoQ 10 and a-tocopherol (54). In the presence of FSP1, CoQ 10 was reduced to CoQ 10 -H 2 using NADPH to inhibit the propagation of PL peroxidation. Inhibition of FSP1 can robustly sensitize cells to RSL3-induced ferroptosis (53, 54).

NRF2
Nuclear factor erythroid 2-related factor 2 (NRF2) encoded by NFE2L2 is a transcription factor that consists of 605 amino acids and contains seven conserved homology domains termed as Neh1-Neh7 (128,129). Neh2 is the major regulatory domain that has two binding motifs known as DLG and ETGE, both of which help to regulate NRF2 stability by interacting with the Kelch domains of the E3 ubiquitin ligase Kelch-like ECHassociated protein 1 (KEAP1), which is the most notable negative regulator of NRF2 and contains a number of cysteine residues (including C151, C273, and C288) that act as electrophilic attack centers (130). Under physiological conditions, NRF2 is maintained at low levels in normal tissues because of KEAP1-dependent ubiquitination (130). KEAP1 acts as an adaptor for NRF2 binding to the KEAP1-cullin 3-ring box protein 1 (KEAP1-CUL3-RBX1) E3 ubiquitin ligase complex, which targets NRF2 for rapid proteasomal degradation (129). During increased oxidative stress, or if KEAP1, CUL3, or NRF2 are mutated, NRF2 can no longer be ubiquitylated and degraded, and the newly translated NRF2 transfers to the nucleus and activates transcription of anti-oxidant response element (ARE)containing genes, many of which play a key role in preventing the initiation of ferroptosis (131). RSL3 or ML-162 treatment increases the expression of p62 and NRF2 in chemo-resistant HN3R and HN3-rslR cells, inactivates KEAP1, and increases the expression of the phospho-protein kinase R-like endoplasmic reticulum kinase (p-PERK)-activating transcription factor 4 (ATF4)-sertrin 2 (SESN2) (132). NRF2 inhibition could reverse the resistance of cisplatin-resistant HNC cells to artesunateinduced ferroptosis (55). With regard to other Neh domains, Neh1 known as a DNA binding domain enhances NRF2 transcriptional activation (133). The Neh3, Neh4, and Neh5 domains are known as trans-activation domains of NRF2 (134)(135)(136). Neh6, a serine-rich domain, negatively modulates NRF2 stability (137), and Neh7 has been shown to interact with the nuclear receptor retinoid X receptor a (RXRa), which inhibits the transcription of NRF2 target genes (138).

p53
The tumor suppressor gene TP53, regarded as the "guardian of the genome", was first discovered in 1979 in a complex with the simian virus 40 large T antigen and has been widely studied in cancers (153). Many studies have suggested that p53 could act synergistically with established oncogenes to promote the conversion of normal tissues to tumors and lead to accelerated metastasis. The first evidence that p53 might act as a tumor suppressor was reported in 1984 (154). Subsequently, it was concluded that these previous experiments were performed not with wild-type p53, but with its mutated version (155). The mutant p53 is highly abundant in cancers, and it promotes tumorigenesis by disabling the function of the wild-type p53, as well as by "gain-of-function" processes such as the accumulation of p53 mutations, that augments the oncogenic capacity of the mutated p53 and delivers a stronger tumoral resistance against anti-cancer treatments (156,157). Under normal conditions, the levels of p53 are low because of the E3 ubiquitin-protein ligase MDM2, which can target and degrade p53, while oncogene activation could prevent MDM2 binding to p53 and stimulate p53 acetylation (158). Although 42% of cases across 12 tumor types occur due to TP53 mutation, the rate varies widely across diverse types of cancers (159).
Apart from the effects of p53 on apoptosis, autophagy, and the cell cycle, it also regulates ferroptosis via transcriptional or post-transcriptional mechanisms. Intriguingly, p53 can enhance or inhibit ferroptosis via different pathways. As to proferroptosis, p53 promotes ferroptosis of MCF7 and U2OS cells due to SLC7A11 repression without affecting the expression of other p53 target genes involved in the cell cycle and apoptosis (56). p53 targets spermidine/spermine N1-acetyltransferase 1 (SAT1), which correlates with the expression of 15-LOX, and promotes the expression of SAT1 to induce ferroptosis (57). Depletion of p53 prevents nuclear accumulation of dipeptidyl peptidase-4 (DPP4) and triggers membrane-associated DPP4mediated lipid peroxidation by binding to NADPH oxidase 1 (NOX1), which culminates in the ferroptosis of colorectal cancer cells. On the other hand, p53 mediates expression of the tumor suppressor cyclin-dependent kinase inhibitor 1A (CDKN1A/ p21), which is a key mediator of p53-dependent cell cycle arrest after DNA damage, delays the onset of ferroptosis in response to cystine deprivation in cancer cells, and increases p53 expression by using the MDM2 inhibitor nutlin-3 that blocks erastin-induced ferroptosis in HT-1080 cells (58).

FERROPTOSIS AND CANCER
Here, different cancer groups are discussed in the context of ferroptosis, and the numerous inducers and inhibitors of ferroptosis, including drugs, genes, and RNA are listed in Table 2.

Pharmaceuticals-Induced Ferroptosis
Many clinical drugs have the ability to induce ferroptosis in cancer cells, which means that researchers need to comprehensively explore the potential mechanisms of these drugs. Artesunate, an anti-malarial drug, can specifically induce ferroptosis in pancreatic cancer cells without affecting pancreatic ductal epithelial cells (169). Artesunate also activates the ATF4-CHOP-CHAC1 cascade in DAUDI and CA-46 cells, and CHAC1 enhances artesunate-induced ferroptosis in Burkitt's lymphoma cells (210). Dihydroartemisinin, a semisynthetic derivative of artemisinin, increases ROS levels in a dose-dependent manner and decreases the levels of both GPX4 and RAS in HEP-2 and CNE-1 cells (189). Dihydroartemisinin can also regulate the activity of the AMP-activated protein kinase (AMPK)/mTOR/p70S6k signaling pathway, thereby accelerating ferritin degradation, increasing LIP, promoting cellular ROS accumulation, and eventually triggering ferroptosis in acute myeloid leukemia cells (200). Several anti-inflammatory drugs have shown the potential to induce ferroptosis. Combinatorial treatment of auranofin and BSO can downregulate GPX4 and accumulate NRF2 and HO-1 (162). Ibuprofen induces ferroptosis in glioblastoma cells by downregulating the NRF2 signaling pathway (213). Acetaminophen can strengthen the sensitivity of erastin-induced ferroptosis by modulating the NRF2/HO-1 signaling pathway in non-small-cell lung cancer (NSCLC) (145). SAS can induce ferroptosis in breast cancer cells by upregulating TFR1 and DMT1, especially in cells with low expression of estrogen receptors (47). The tranquilizer haloperidol promotes ferroptosis by accelerating iron accumulation, lipid peroxidation, and GSH depletion under erastin or sorafenib treatment in HepG2 and Huh-7 cells (160). The lysosome disruptor siramesine and the tyrosine kinase inhibitor lapatinib synergistically induce ferroptosis by increasing the expression of transferrin and decreasing the expression of FPN (188).
Many extracts from plants and herbs also exhibit anti-tumor effects by inducing ferroptosis. Ruscogenin can induce ferroptosis by improving ROS generation and trigger intracellular iron accumulation by upregulating transferrin and downregulating FPN (49). Piperlongumine can also induce ferroptosis in pancreatic cancer cells (171). Amentoflavone represses the expression of FTH by autophagy via activation of the AMPK/ mTOR/p70S6K signaling pathway to trigger in vitro and in vivo ferroptosis in an autophagy-dependent manner (177). Typhaneoside, a major flavonoid found in the extract of Pollen Typhae, promotes the activation of the AMPK signaling pathway to contribute to ferritin degradation, ROS accumulation, and ferroptosis in Kas-1, HL-60, and NB4 cells (202). Pseudolaric acid B isolated from cortex pseudolaricis triggers ferroptosis in vivo and in vitro by upregulating TFR1, activating NOX4, and inhibiting SLC7A11 (178). Actinidia chinensis Planch (ACP) has been shown to increase the accumulation of ROS by inhibiting GPX4 and SLC7A11 to induce ferroptosis in HGC-27 cells (193). Dihydroisotanshinone I, a pure compound present in danshen, can induce ferroptosis by downregulating the expression of GPX4 both in vitro and in vivo (186). The extract of Betula etnensis Raf.
Some new compounds are still being found by researchers. The novel compound ferroptocide, identified as an inhibitor of thioredoxin, could rapidly and robustly induce ferroptotic cell death and positively modulate the immune system in a breast cancer model (187). Phenethyl isothiocyanate, present in cruciferous vegetables, induces ferroptosis, autophagy, and apoptosis in K7M2 cells by activating the ROS-related mitogen-activated protein kinase (MAPK) signaling pathway (211).

Positive Regulators
Overexpression of some genes and proteins can promote ferroptosis in cancer cells. The level of DNAJ/Hsp40 homolog subfamily B member 6 (DNAJB6) is negatively correlated with lymph node metastasis in esophageal squamous cell carcinoma (ESCC) patients, and the overexpression of its isoform DNAJB6a is accompanied by remarkable reduction in the protein levels of GPX4 and phospho-AKT (p-AKT), thus DNAJB6a plays an antioncogenic role in ESCC progression via ferroptosis (212). Signal transducer and activator of transcription-3 (STAT3) can promote ferroptosis through activation of cathepsin Bmediated lysosomal cell death in PANC-1 and CFPAC-1 cells (170). Loss of function of genes and proteins could change the sensitivity of cancer cells to ferroptosis. The loss of function of retinoblastoma (Rb) protein is significant during liver carcinogenesis, and the Rb-negative status of hepatocellular carcinoma (HCC) cells promotes ferroptosis on sorafenib treatment (163). Knockdown of high mobility group box 1 (HMGB1) decreases erastin-induced ROS generation via an iron-mediated lysosomal pathway in HL-60 cells expressing NRAS Q61L (201).
Some genes and proteins can affect lipid metabolism and subsequently induce ferroptosis. Bromelain induces ferroptosis by inhibiting the proliferation of KRAS mutant colorectal cancer cells via ACSL4 (192). ADP-ribosylation factor 6 (ARF6) can sensitize gemcitabine-resistance pancreatic cancer cells to RSL3induced lipid peroxidation by affecting the ACSL4 protein levels (172). Hypoxia-inducible transcriptional factor 2a (HIF-2a) selectively enriches PUFAs by activating the expression of the hypoxia-inducible lipid droplet-associated (HILPDA) protein and induces a ferroptosis-susceptible cell state in 786-O cells (208).
Several cellular signaling pathways are also involved in ferroptosis. Removal of the Hippo pathway effector transcription regulator 1 (TAZ) confers ferroptosis resistance, whereas overexpression of TAZS89A sensitizes cells to ferroptosis. TAZ promotes ferroptosis by regulating epithelial membrane protein 1 (EMP1) and NOX4 in renal cell carcinoma (207). Low levels of TAZ in chemo-resistant recurrent ovarian cancer are responsible for reduced ferroptosis susceptibility (203). Blockade of MAPK signaling similarly protects cells from ferroptosis, whereas NSCLC with sustained MAPK activation is likely to respond to ferroptosis following cystine depletion (206).

Negative Regulators
Many genes and proteins can protect cancer cells from ferroptosis. The sigma-1 receptor (S1R) protects HCC cells against sorafenib, while knockdown of S1R could induce ferroptosis by blocking the expression of GPX4 (164). The human mitochondrial cysteine desulfurase (NFS1) is highly expressed in well-differentiated lung adenocarcinoma and protects cells from ferroptosis (182). CDGSH iron sulfur domain 1 (CISD1), an iron-containing outer mitochondrial membrane protein, negatively regulates erastin-induced ferroptosis in HepG2 and Hep3B cells by inhibition of mitochondrial lipid peroxidation in a GPX4-independent manner (168). Overexpression of the CISD2 gene has been shown to confer resistance to HN-6 and HN-10 cells against SAS-induced ferroptosis, while silencing CISD2 could reverse SAS-resistant cells to a ferroptotic state with increased levels of lipid ROS and mitochondrial Fe 2+ (190).
Some genes and proteins hinder ferroptosis; however, ferroptosis can occur when these are suppressed. NRF2 can upregulate cystathionine b-synthase (CBS) to confer resistance to erastin-induced ferroptosis in ovarian cancer cells (204), however inhibition of CBS can trigger ferroptosis in HCC (167). Inhibition of Lon peptidase 1 (LONP1) could contribute to ferroptosis of PANC-1 cells by activating the NRF2/KEAP1 signaling pathway and upregulating GPX4 expression (175). Heat shock protein family A member 5 (HSPA5) can upregulate the expression and activity of GPX4 to prevent dihydroartemisinin-induced ferroptosis in glioma (176). Inhibition of glucose-regulated protein 78 (GRP78), one of the most active molecular chaperone components in the ER of cancer cells, could enhance the effect of artesunate-induced ferroptosis in KRAS mutant pancreatic cancer cells (174). Growth/differentiation factor 15 (GDF15) knockdown has been shown to promote erastin-induced ferroptosis in MGC803 cells by attenuating SLC7A11 expression and subsequently decreasing intracellular GSH levels (179).
Some genes and proteins can prevent ferroptosis via affecting lipid metabolism. EGLN1/c-Myc directly activates the expression of lymphoid-specific helicase (LSH) by inhibiting HIF-1a, then LSH as an oncogene epigenetically increases the expression level of lipid metabolic genes that inhibit ferroptosis in lung cancer (183). Reconstitution of the functional vov Hippel-Lindau (VHL) gene prevents ferroptosis by reverting cells back to an oxidative metabolism and increasing fatty acid degradation through boxidation (209). Steroyl-CoA desaturase 1 (SCD1), the ratelimiting enzyme of MUFA synthesis, is highly expressed in ovarian cancer tissues and cell lines, and overexpression of SCD1 could protect cells from ferroptosis (205).
Several genes and proteins related to ferroptosis can be utilized as prognostic markers. G12D is the most frequent mutation in KRAS (referred as KRAS G12D ), and it is confirmed that KRAS G12D could be released from PDAC cells succumbing to autophagy-dependent ferroptosis, and then KRAS G12D causes macrophages to switch to an M2-like pro-tumor phenotype via STAT3-dependent fatty acid oxidation and it is also found that high KRAS G12D expression in macrophages is correlated with poor survival in PDAC patients (173). High expression of Serine Threonine Tyrosine kinase 1 (STYK1) predicts poorer prognosis and is related to high levels of GPX4 in NSCLC (184).
A few long non-coding RNAs (lncRNAs) may also regulate ferroptosis. Erastin could upregulate the lncRNA GABPB1-AS1, which could downregulate the level of GABPB1 by blocking GABPB1 translation, leading to the downregulation of peroxiredoxin-5 and suppression of the cellular anti-oxidant capacity, and high expression of GABPB1 has been correlated with poor prognosis in HCC patients (166). The lncRNA LINC00336 is upregulated in lung cancer and functions as an inhibitor of ferroptosis in carcinogenesis by interacting with ELAVlike RNA-binding protein 1, which acts as a novel regulator of ferroptosis (185).

CONCLUSION AND PERSPECTIVE
In recent years, numerous studies have been conducted to study ferroptosis in cancers, and several anti-tumor drugs (such as sorafenib, SAS, altertamin) have been approved by the FDA. However, many unsolved issues regarding ferroptosis still exist. What is the executor of ferroptosis in cancer? Which is the main pathway of lipid peroxidation in ferroptosis, the nonenzymatic pathway or enzymatic one? Does there exist another mechanism of ferroptosis resistance? What is the likelihood that the ferroptosis-related anti-tumor drugs could be used in clinical settings, such as post-operation, chemo-resistance, or radioresistance instances? What are the potential adverse effects of ferroptosis-related anti-tumor drugs? Could the administration of ferroptosis-related anti-tumor drugs be combined with immunotherapy? The research on ferroptosis is still in its infancy, and more studies are needed to comprehensively explore the mechanism of ferroptosis in various cancers.

AUTHOR CONTRIBUTIONS
ZS and CS designed and conceptualized the review. All authors contributed to the article and approved the submitted version.