Ferroptosis and Cancer: Mitochondria Meet the “Iron Maiden” Cell Death

Ferroptosis is a new type of oxidative regulated cell death (RCD) driven by iron-dependent lipid peroxidation. As major sites of iron utilization and master regulators of oxidative metabolism, mitochondria are the main source of reactive oxygen species (ROS) and, thus, play a role in this type of RCD. Ferroptosis is, indeed, associated with severe damage in mitochondrial morphology, bioenergetics, and metabolism. Furthermore, dysregulation of mitochondrial metabolism is considered a biochemical feature of neurodegenerative diseases linked to ferroptosis. Whether mitochondrial dysfunction can, per se, initiate ferroptosis and whether mitochondrial function in ferroptosis is context-dependent are still under debate. Cancer cells accumulate high levels of iron and ROS to promote their metabolic activity and growth. Of note, cancer cell metabolic rewiring is often associated with acquired sensitivity to ferroptosis. This strongly suggests that ferroptosis may act as an adaptive response to metabolic imbalance and, thus, may constitute a new promising way to eradicate malignant cells. Here, we review the current literature on the role of mitochondria in ferroptosis, and we discuss opportunities to potentially use mitochondria-mediated ferroptosis as a new strategy for cancer therapy.


Introduction
Ferroptosis is a non-apoptotic, iron-dependent form of regulated cell death (RCD) occurring when the intracellular levels of lipid reactive oxygen species (L-ROS) exceed the antioxidant activity of glutathione-dependent peroxidase (GPX4) thus leading to the collapse of cellular redox homeostasis [1]. Ferroptosis is defined by three essential hallmarks: (i) oxidation of polyunsaturated fatty acid (PUFA)-containing membrane phospholipids; (ii) availability of redox-active iron; and (iii) loss of lipid hydroperoxide (LOOH) repair capacity [2]. The physiological function of ferroptosis as well as its involvement in multiple human diseases, such as ischemic organ injury, neurodegeneration, and cancer, have been established [3][4][5][6].

The Role of Iron Metabolism in Ferroptosis
Given its unique redox properties, iron is often incorporated as a prosthetic group in enzymes and structural proteins and participates in many enzymatic reactions, thus representing a key player in many cellular biological processes [16]. The same features make iron potentially dangerous, as it can donate electrons to O 2 and H 2 O 2 to generate potentially harmful ROS such as hydroxyl radicals, hydroperoxyl radicals, and superoxide anions [53]. To ensure both fulfillment of metabolic needs and minimization of toxicity, cells are provided a complex protein network that tightly regulates iron import, storage, and detoxification ( Figure 2) [62].
Briefly, transferrin (TF) imports circulating iron (Fe 3+ ) into the cell by binding to its specific receptor TFR1 [63]. Once internalized, the TF-Fe 3+ -TFR1 complex localizes into the endosomes; here, in response to acidic conditions, iron is reduced to Fe 2+ by the six-transmembrane epithelial antigen of the prostate 3 (STEAP3) and then exported into cytosol by divalent metal transporter 1 (DMT1) [64]. The majority of cytoplasmic iron is stored within ferritin, a nanocage composed of 24 subunits of both light (ferritin light chain, FtL) and heavy (ferritin heavy chain, FtH) types [65,66]. Ferritin heavy chain, in particular, is provided with ferroxidase activity through which it maintains iron in its ferric Fe 3+ non-toxic form [67,68]. A small pool of cytoplasmic free Fe 2+ , referred to as LIP, directly catalyzes free radical formation via Fenton Reaction [69]. Excess Fe 2+ is then oxidized to Fe 3+ and exported by ferroportin (FPN) [70].
The expression of TFR1 and FtH is regulated by the interaction between the iron regulatory proteins (IRPs) and the iron-responsive element (IRE), a stem-loop structure located in the 3' UTR of FtH mRNA and in the 5 UTR of TFR1 mRNA. In response to cellular iron demand IRE/IRP interaction promotes TFR1 mRNA stability and inhibits FtH translation, thus modulating cellular iron uptake and storage [71].
Overexpression of both TF and TFR1 sensitizes cells to ferroptosis by enhancing iron uptake [72]; on the contrary, silencing TFR1 can inhibit erastin-induced ferroptosis. In this regard, it has been recently demonstrated that heat shock protein beta-1 (HSPB1) significantly inhibits ferroptosis by repressing TRF1 expression and, thus, reducing intracellular iron concentrations [38]. These proteins are a family of highly conserved molecular chaperones that, once activated by environmental stress, promote cell resistance to different types of cell death-including ferroptosis [73]. Heat shock protein family A member 5 (HSPA5), an endoplasmic reticulum (ER)-sessile chaperone, binds and stabilizes GPX4, thus indirectly counteracting lipid peroxidation in ferroptosis [73,74]. Iron crossroads from cytosol to mitochondria. Cytosolic iron metabolism: 1) TFR1 internalizes Fe 3+ -loaded TF through an endocytosis-mediated mechanism. 2) Fe 2+ uptake is carried out by the transmembrane permeable channel DMT1. 3) NTBI enters cytoplasm through the zinc transporter ZIP 8/14 upon its reduction in Fe 2+ mediated by PRNP. 4) Fe 3+ -loaded TF and NTBI are released in the endosome by TFR1 and ZIP8/14, respectively. STEAP3 converts Fe 3+ to Fe 2+ which, in turn, enters the cytoplasm via DMT1. After internalization, all these carriers are recycled to the cell surface. 5) GRX3 and BOLA2 constitute a heterotrimeric complex involved in the CIA system for (Fe-S) cluster formation. 6) PCBP1/2 iron chaperones bind iron and deliver it via direct protein-protein interaction with PHD2, FIH1, DOHH, and ferritin, in a process known as metallation. 7) LIP is a pool Figure 2. Iron crossroads from cytosol to mitochondria. Cytosolic iron metabolism: (1) TFR1 internalizes Fe 3+ -loaded TF through an endocytosis-mediated mechanism. (2) Fe 2+ uptake is carried out by the transmembrane permeable channel DMT1. (3) NTBI enters cytoplasm through the zinc transporter ZIP 8/14 upon its reduction in Fe 2+ mediated by PRNP. (4) Fe 3+ -loaded TF and NTBI are released in the endosome by TFR1 and ZIP8/14, respectively. STEAP3 converts Fe 3+ to Fe 2+ which, in turn, enters the cytoplasm via DMT1. After internalization, all these carriers are recycled to the cell surface. (5) GRX3 and BOLA2 constitute a heterotrimeric complex involved in the CIA system for (Fe-S) cluster formation. (6) PCBP1/2 iron chaperones bind iron and deliver it via direct protein-protein interaction with PHD2, FIH1, DOHH, and ferritin, in a process known as metallation. (7) LIP is a pool of free and redox-active iron which promotes ROS generation through a Fenton Reaction. (8) Ferritin is an iron-storage protein with ferroxidase activity, able to convert toxic Fe 2+ in non-toxic Fe 3+ , thus preventing a Fenton Reaction. Cells 2020, 9, 1505 7 of 26 (9) IRPs coordinate iron homeostasis at the post-transcriptional level. IRP1/2 blocks degradation of TFR1 mRNA and inhibits the translation of both ferritin subunits, FtH and FtL, and FPN. (10) FPN exports iron in the extracellular space; its activity is decreased by hepcidin that directly binds to FPN. Mitochondrial iron metabolism: (11) LIP released by lysosomes is rapidly taken up by MCU and internalized into mitochondria. (12) Mfrn1/2 imports Fe 2+ from the intermembrane space of the mitochondria to the mitochondrial matrix. (13) VDAC2/3 mediates iron mitochondrial uptake. (14) Endosomal iron is delivered in mitochondria through the so-called "kiss and run" mechanism. (15) FECH forms an oligomeric complex with ABCB10 to synergistically promote mitochondrial iron import. (16) Fe 2+ participates to Fenton Reaction-generating mitoROS. (17) FtMt, an H-type ferritin, is involved in mitochondrial iron storage. (18) PPIX incorporates iron to generate heme and mediates ISC export. (19) Mitochondrial iron can even enter the ISC assembly machinery, responsible for the maturation of all cellular (Fe-S) clusters; then, it can be mobilized to OXPHOS complex I/II/III. (20) NEET iron-sulfur proteins transfer their 2Fe-2S clusters to an apo-acceptor protein and CIA system. Changes of ferritin expression levels affect ferroptosis by altering the intracellular free and redox active iron pool. Torii et al. [75] have demonstrated that NCOA4 overexpression reinforces ferritin degradation and then drives ferroptosis, while NCOA4 knockdown suppresses ferritin degradation and inhibits ferroptosis. Increased expression of ferritin restrains the expansion of LIP and limits ferroptosis [76,77]. Indeed, suppression of IRE-binding protein IREB2, through RNA interference, significantly increases the expression of FtL and FtH subunits, thereby limiting erastin-induced ferroptosis [46].
Taken together, these data clearly indicate that the imbalance of intracellular iron homeostasis in favor of iron overload is pivotal for the induction of ferroptosis.

Ferroptosis and Cancer
Acting as an adaptive mechanism to eliminate malignant cells, ferroptosis constitutes a new tumor suppressing pathway [35]. Although initially defined as a new form of cell death occurring in RAS mutant cancer cells, it is now clear that the RAS pathway is not the sole determinant of ferroptosis occurrence in tumor [78,79].
Biochemically, two central events, intracellular iron accumulation and lipid peroxidation, are required for ferroptosis fulfillment in cancer cells [80]. The metabolite-mediated ways for inducing ferroptosis include decreasing cystine uptake through the inhibition of system x c − and targeting GPX4, therefore, increasing iron concentration and ROS [81,82]. The tumor suppressor p53 plays a role in both inhibition and promotion of ferroptosis, depending on the cellular context. It can induce ferroptosis by inhibiting the transcription of the SLC7A11 gene encoding the substrate-specific subunit of system x c − [2,82]. Repression of SLC7A11, blocks cystine uptake and suppresses GPX4 activity, thus rendering cancer cells prone to undergo ferroptosis upon oxidant insults [83]. In this regard, in vivo studies have demonstrated that, while acetylation-defective p53 mutant (TP53-3KR) fails to trigger cell senescence, apoptosis and cell-cycle arrest, it is still able to suppress tumorigenesis via ferroptosis [84]. On the other hand, p53 is provided with an anti-ferroptotic function related to its capacity to boost antioxidant defense. This activity is most likely mediated by the p53/p21 axis activation that, preserving GSH and other thiols, suppresses phospholipid oxidation [85]. These observations are in line with the ability of p53 to limit erastin-induced ferroptosis in colorectal cancer (CRC) cells [86]. Ferroptosis is also promoted by the activity of p53 involvement in mevalonate pathway which generates a series of metabolites, including squalene and ubiquinone, with potential anti-ferroptotic activity [87]. When metabolic stress conditions occur, p53 promotes the expression of ATP-binding cassette subfamily A member 1 (ABCA1) that, in turn, regulates cholesterol efflux from the plasma membrane to the endoplasmic reticulum, causing inhibition of sterol regulatory element binding protein 2 (SREBP2) [88]. Inactivation of SREBP2 alters the mevalonate pathway, preventing the production of squalene and ubiquinone [89].
The tumor suppressor p53 can also induce ferroptosis by activating lipoxygenase ALOX12 function. Briefly, the transcriptional repression of SLC7A11 leads to ALOX12-dependent ferroptosis upon oxidative stress [90]. Other lipoxygenases, including ALOXE3 and ALOX15B, are essential for ferroptosis occurrence in cancer. A comprehensive study showed that erastin-induced ferroptosis is rescued by silencing either ALOX15B or ALOXE3 in transformed fibroblasts (BJeLR) and fibrosarcoma (HT-1080) cells [42]. Epigenetic regulation also plays a key role in ferroptosis. Loss of function mutations of the tumor suppressor BRCA1-associated protein 1 (BAP1) has recently been linked to ferroptosis [91]. This BAP1, a nuclear-located deubiquitinase (DUB), promotes the formation of the polycomb-repressive-deubiquitinase (PR-DUB) complex and reduces histone 2A ubiquitination (H2Aub) on the SLC7A11 promoter [92]. The consequent downregulation of SLC7A11 blocks ferroptosis, as it leads to cystine starvation and depletion of GSH [93].
A plethora of long non-coding RNAs (lncRNAs) and microRNAs (miRNAs) have recently been reported to regulate ROS metabolism and ferroptosis [94,95]. For example, P53RRA lncRNA promotes cell-cycle arrest, apoptosis, and ferroptosis by binding to Ras GTPase-activating protein-binding protein 1 (G3BP1) and preventing its interaction with p53 which is consequently retained in the cell nucleus [96]. The microRNA 137 inhibits ferroptosis by targeting SLC1A5 which results in dysfunction of the glutamine transporter in cancer cells. The metabolism of L-Gln contributes to the formation of oxidizable lipids to induce ferroptosis. The Gln importer SLC1A5/SLC38A1, glutaminases 2 (GLS2), and glutamic-oxaloacetic transaminase 1 (GOT1) are required for Gln uptake and metabolism to Glu and ultimately to α-ketoglutarate (α-KG). Accordingly, miR-137 overexpression suppresses erastin/RSL3-induced ferroptosis in melanoma cells [97].
To sum up, numerous genes/proteins and metabolic pathways are involved in the execution of ferroptosis in cancer cells. A more extensive list of the main molecules implicated in ferroptosis is reported in Table 1.
Morphologically, ferroptotic cancer cells exhibit alterations of mitochondrial morphology and cristae structure. Upon treatment with erastin in vitro, ferroptotic BJeLR cancer cells are usually rounded up and detached [45]. Transmission electron microscopy (TEM) reveals the presence of small mitochondria with increased mitochondrial membrane density and vanishing of mitochondrial cristae. The cell membrane remains intact and the nucleus shows a normal size, without chromatin concentration [98,99]. This table lists several genes, proteins, microRNAs, and lncRNAs, acting as modulators of the ferroptosis process. The molecules are divided according to their pro-or anti-ferroptotic action. The molecular mechanism through which these molecules act is also illustrated.

Ferroptosis Is a New Promising Target for Cancer Treatment
Metabolic reprogramming leads to the acquisition of ferroptosis sensitivity as part of an escape strategy against other therapies [112]. This observation strongly supports the potential use of ferroptosis initiating therapies (FITs) in the management of the so-called "persister cells", a subset of cancer cells able to survive upon treatment with several rounds of chemotherapy drugs, leading to tumor relapse [113].
Several drugs targeting ferroptosis have been tested as a new approach in anti-tumor therapies in vitro. Overall, these drugs can be classified as follow: (i) drugs directly or indirectly inhibiting system x c − (i.e., erastin, sorafenib, and sulfasalazine); (ii) drugs inhibiting GSH synthesis through the suppression of γ-glutamylcysteine synthetase (GCS) (i.e., buthionine sulfoximine, BSO); and (iii) drugs inhibiting GPX4 (i.e., RSL3, withaferin A and FIN56). Among the abovementioned drugs RSL3 and erastin, the two main ferroptosis inducers have been used in a variety of tumor models in vitro [114,115]. Both drugs, however, do not meet the pharmacokinetic standards for in vivo application and need to be further optimized for clinical application [116]. Sorafenib is an FDA-approved multi-kinase inhibitor for treatment of advanced renal cell carcinoma (RCC) and advanced HCC. Sulfasalazine (SSZ) disrupts iron metabolism through cystine uptake blockade, resulting in ferroptosis of glioma cells [117].
A drug screening analysis has indicated that some of the well-known chemotherapeutics, such as cisplatin, altretamine, and artesunate, are able to promote ferroptosis [118]. Cisplatin leads to GSH depletion and GPX4 inactivation [119]; indeed, it is emerging as inducer of both ferroptosis and apoptosis in A549 non-small cell lung cancer (NSCLC) cells and HCT116 CRC cells [119]. Altretamine (hexamethylmelamine), an FDA-approved alkylating antineoplastic drug used for treating ovarian cancer, inhibits GPX4 and effectively kills U-2932 diffuse large B cell lymphoma (DLBCL) cells in vitro [120]. Artesunate (ART) increases ROS generation in cancer cells, and its antitumor effect is carried out through ferroptosis in a variety of neoplastic diseases like pancreatic ductal adenocarcinoma (PDAC), epithelial ovarian cancer (EOC), and HNCs [39].
Recently, nanocarriers have been proposed as an efficient approach to induce ferroptosis in cancer cells in vitro and in vivo [121]. Doxorubicin, packed into mesoporous carbon nanoparticles, induces ferroptosis in breast cancer (MCF7) cells, in A549 cells, and in human cervical carcinoma (HeLa) cells [122]. The nano-targeting of withaferin A, a natural ferroptosis-inducing agent, efficiently kills high-risk neuroblastoma cell lines and suppresses growth of neuroblastoma xenografts in mice [123]. Iron-based nanoparticles can release Fe 2+ and Fe 3+ in acidic lysosomes, inducing ferroptosis, ultimately suppressing tumor growth [124]. In this regard, treatment with small (∼6 nm) surface-functionalized poly(ethylene glycol)-coated (PEGylated) silica nanoparticles (C' dots) can induce ferroptosis in tumor xenografts by delivering iron into cells [125]. Since exogenous iron overload (e.g., ferric ammonium citrate) is not able to induce ferroptosis in all cell types, it is likely that treatment with C' dots is effective in a cell-specific manner [2]. Numerous other molecules have been shown to trigger ferroptotic cell death in cancer cells and future studies will be necessary for their validation in real clinical settings.
The array of compounds able to induce ferroptosis in cancer cells is summarized in Table 2. Nonetheless, for a more detailed discussion about drugs currently used in ferroptosis-based cancer treatment, we recommend the exhaustive review by Bin Lu et al. [11].  [137] iron ionophores sequestration of iron into lysosomes and stimulation of ferritin degradation [138] poly(butylcyanoacrylate) and zero-valent iron nanoparticles and arginine-rich manganese silicate nanobubbles induction of oxidative stress and lipid peroxidation [139] Nanocarriers (doxorubicin into mesoporous carbon nanoparticles/withaferin A/poly(ethylene glycol)-coated (PEGylated) silica nanoparticles (C' dots ) induction of oxidative stress [122,123,125] This table summarizes the available chemotherapeutic agents and targeted compounds able to induce or inhibit ferroptosis. Specific targets and mechanisms are also reported.

Mitochondria at the Crossroad of Ferroptosis and Cancer Suppression
Mitochondria play a pivotal role in metabolic plasticity in malignant cells, as well as in the regulation of many RCD processes, and ferroptosis is no exception [140]. Mitochondria seem to be involved in ferroptosis induced by cystine deprivation (CDI) which, indeed, is associated with mitochondrial membrane hyperpolarization and lipid peroxide accumulation [26]. In agreement, erastin treatment boosts the production of mitoROS [26] which, in turn, cause opening of mitochondrial permeability transition pore (mPTP), dissipation of ∆Ψ m and ATP depletion [141]. Cells undergoing ferroptosis exhibit mitochondria fragmentation and specific changes in mitochondrial morphology such as reduction of mitochondrial cristae and decrease in mitochondrial size [142].
However, some questions remain controversial. Whether mitochondrial dysregulation is able, per se, to initiate this type of cell death or it is just a consequence of the metabolic imbalance is unclear. Based on Gaschler et al. [134], cells lacking mitochondria are still sensitive to ferroptosis. Conversely, according to Gao et al. [26], inhibition of TCA cycle and mitochondrial ETC can rescue cells from mitochondrial membrane hyperpolarization, lipid peroxide accumulation, and ferroptosis. Mitochondrial role in ferroptosis seems context dependent. Upon cystine deprivation, mitochondria contribute to reducing GSH and to promoting ROS production. Glutaminolysis is required for CDI ferroptosis.
In the following sections, we review the morphological, metabolic, and energetic features that closely relate mitochondria to ferroptotic cell death (Figure 3). Cells undergoing ferroptosis exhibit mitochondria fragmentation and specific changes in mitochondrial morphology such as reduction of mitochondrial cristae and decrease in mitochondrial size [142]. However, some questions remain controversial. Whether mitochondrial dysregulation is able, per se, to initiate this type of cell death or it is just a consequence of the metabolic imbalance is unclear. Based on Gaschler et al. [134], cells lacking mitochondria are still sensitive to ferroptosis. Conversely, according to Gao et al. [26], inhibition of TCA cycle and mitochondrial ETC can rescue cells from mitochondrial membrane hyperpolarization, lipid peroxide accumulation, and ferroptosis. Mitochondrial role in ferroptosis seems context dependent. Upon cystine deprivation, mitochondria contribute to reducing GSH and to promoting ROS production. Glutaminolysis is required for CDI ferroptosis.
In the following sections, we review the morphological, metabolic, and energetic features that closely relate mitochondria to ferroptotic cell death (Figure 3).

Mitochondrial Morphological Features in Ferroptosis
The ultrastructural changes of mitochondria are considered the morphological trademark of ferroptosis that help to distinguish this new type of RCD from apoptosis, necroptosis, and autophagy [99]. These changes occur upon both pharmacological and genetical induction of ferroptosis in all cell types. Considering that biomarkers exclusively associated with ferroptosis are missing, the detection of typical mitochondrial morphological changes by TEM represents one of the few available methods for the identification of ferroptosis [98]. A list of the available methods for the in-depth characterization of mitochondrial function in ferroptosis is reported in Table 3. Table 3. Main methods to characterize mitochondrial function in ferroptosis.

Biological Context Reagents Functions References
Morphological changes TEM detects ultrastructural mitochondrial morphology changes in the occurrence of ferroptosis [142] Mitochondrial oxidative stress MitoSOX detects mitochondrial superoxide formation in live cells [149] MitoTEMPO mitochondrially targeted antioxidant, a specific scavenger of mitochondrial superoxide; it can be used in combination with MitoSOX reagent as positive control [150] Mitotracker fluorescent dye that stains mitochondria in live cells and its accumulation is dependent upon membrane potential; in can be also used coupled with MitoSOX, in order to stain mitochondrial superoxide and mitochondria together [26] Lipid peroxidation BODIPY detects reactive oxygen species generated by lipid peroxidation in mitochondrial and plasma membranes using flow cytometry [151] ∆Ψm TMRE quantifies changes in mitochondrial transmembrane potential (∆Ψm) in live cells by flow cytometry, microplate spectrophotometry and fluorescent microscopy [152] This table summarizes the available methods and reagents used to explore the pivotal mechanisms and alterations involving mitochondrial function in ferroptosis. Biological context and specific functions are also illustrated for each reported method.
The morphological features of ferroptotic cells can be classified based on the extent of mitochondria fragmentation and their distribution: (i) uniformly distributed, elongated mitochondria, (ii) uniformly distributed, fragmented mitochondria, (iii) fragmented mitochondria mainly distributed close to the nucleus, (iv) small rounded mitochondria located close to the nucleus [27,153,154]. As previously reported, shrinkage of mitochondria with enhanced mitochondrial membrane density, volume reduction, and vanishing of mitochondrial cristae have been observed in ferroptosis following erastin treatment in BJeLR cells [45]. Induction of ferroptosis by GPX4 knockdown in immortalized fibroblasts and kidney tissue-derived cells has been associated with OMM rupture as observed using TEM. In mouse embryonic fibroblast (MEF) cells (Pfa1 cells), RSL3 treatment induces OMM rupture in a time-dependent manner [116].
In contrast, no morphological features related to necrosis (cytoplasmic swelling, plasma membrane rupture), apoptosis (chromatin condensation and apoptotic bodies) or autophagy (formation of double-membrane enclosed vesicles) were observed following erastin treatment in cancer cells [34,36].

Mitochondrial Energetic Metabolism in Ferroptosis
Mitochondrial metabolism and ferroptosis closely interact with one another. In cystine-deprivation conditions, mitochondrial metabolism significantly contributes to L-ROS generation and ferroptosis [26]. As such, mitochondrial damage and mitoROS production occur upon inhibition of xCT or upon cystine starvation, but are not required for ferroptosis induced by GPX4 inhibition [155].
Concerning the role of glutaminolysis, it has been reported that conditions of cystine deprivation promote mitochondrial respiration and the rapid depletion of GSH, thus inducing ROS accumulation, lipid peroxidation and ferroptosis. In the absence of Gln, neither cysteine starvation nor erastin inhibition of system x c − can induce ferroptosis [26]. Glutaminases 1 (GLS1) and GLS2 catalyze the conversion of Gln into Glu [156]. Of note, ferroptosis can be prevented by pharmacological or the genetic inhibition of mitochondrial isoform GLS2, while the cytoplasmic isoform GLS1 is not able to block this type of RCD [50]. Moreover, GLS2 is a transcriptional target of p53 and is up-regulated during p53-dependent ferroptosis [157]. Transaminases convert Glu into α-KG through the transamination process [158]. Both treatment with the transaminases inhibitor aminooxyacetic acid (AOA) and knockdown of the transaminase GOT1 inhibit CDI ferroptosis in MEFs [34,158]. Blockade of glutaminolysis can be counteracted by supplying TCA cycle with metabolites such as αKG, succinate, fumarate, and malate. These intermediates, all downstream of glutaminolysis, can replace the role of Gln in L-ROS accumulation in both MEFs and HT-1080 cells [26], thus supporting the involvement of the TCA cycle in CDI ferroptosis.
Several enzymes of the TCA cycle (i.e., fumarate hydratase, FH, aconitase, ACO, and citrate synthase CS) are necessary for ferroptosis triggered by cystine starvation or by erastin treatment [26]. Accordingly, inhibition of the TCA cycle mitigates ∆Ψ m hyperpolarization, lipid peroxide accumulation, and ferroptosis [49]. The knockdown of dihydrolipoamide dehydrogenase (DLD), a component of α-KG dehydrogenase complex, blocks the increase of L-ROS amount and ∆Ψ m caused by cystine deprivation-or sulfasalazine treatment-induced ferroptosis in HNC [159]. Loss of FH, which also exerts a tumor suppressor function, confers resistance to CDI ferroptosis in renal cancer cells [26].
The TCA cycle supports electron transport activity of protein complexes located in the inner mitochondrial membrane (IMM). As shown by Gao et al. [26], inhibition of the ETC mitochondrial complex I, complex II, complex III, and complex IV suppresses L-ROS accumulation and ferroptosis induced by cystine starvation or erastin treatment in HT-1080 cells. As master regulator of OXPHOS, mitochondria are the major source of ROS [160]. Indeed, cells with disrupted glycolysis are vulnerable to ferroptosis by rewiring cell metabolism to OXPHOS [161].
Mitochondrial fatty acids metabolism represents an important source for lipid peroxides production during ferroptosis [155]. Increase of the proton conductance of the IMM, ETC inhibition and mPTP opening constitute three of the main mechanisms through which fatty-acid metabolism modulates mitochondrial energy to provoke lipid oxidation [45]. Both ACSF2 and CS regulate synthesis and activation of fatty acids: in detail, ACSF2 forms an activating thioester bond between the fatty acid and CoA, while CS catalyzes the first reaction of the TCA, condensing acetyl-CoA and oxaloacetate to form citrate [45]. Dixon et al. [34] demonstrate that CS and ACSF2 knockdown, in both HT-1080 and BJeLR cells, blocks erastin-induced ferroptosis.
Overall, these results strongly support the role of mitochondrial metabolism in CDI ferroptosis and corroborate the hypothesis that ferroptosis acts as a tumor suppressive mechanism potentially useful for cancer therapeutic approaches.

Mitochondria and Iron Metabolism
Iron is the most prevalent metal inside the mitochondria and actively participates to the physiological functions of these organelles [162,163]. Once imported into the cell, iron can be delivered to mitochondria by several mechanisms including (i) the transient interaction between the transferrin-bound iron within endosomes and the OMM, the so-called "kiss and run" model; (ii) the uptake of low/high molecular weight iron complexes from the cytosolic LIP; and (iii) the transfer of iron bound to metallochaperones such as the poly(C)-binding proteins (PCBPs) (Figure 2) [164,165].
Since mitochondrial iron metabolism mainly occurs in the mitochondrial matrix, iron must cross both the OMM and IMM. Iron transport across the IMM is an active process dependent on the membrane transporter mitoferrin 1 (Mfrn1) and its homolog mitoferrin 2 (Mfrn2) (Figure 2). Dysregulation of Mfrn1/2 leads to mitochondrial iron accumulation and oxidative damage [166]. Of note, recent studies highlight that Mfrn1/2 is impaired in neurological diseases, such as Alzheimer's disease, Huntington's disease, Friedreich's ataxia (FRDA), and Parkinson's disease, which are all linked to ferroptosis [167].
The voltage-dependent anion channels (VDACs), located in the OMM, also regulate the influx of iron in mitochondria ( Figure 2) [168]. Erastin treatment induces VDAC2/3 opening and is associated with mitochondrial iron accumulation and iron-dependent ferroptosis [169,170].
Following import, mitochondrial iron primarily acts as a cofactor in Fe-S cluster-containing proteins (i.e., NADH:ubiquinone oxidoreductase) and heme-containing proteins (i.e., cytochrome c, cytochrome c oxidase, and succinate dehydrogenase) all of which are components of the IMM complexes of the ETC [171]. The biogenesis of Fe-S cluster is driven by the activation of the mitochondrial protein frataxin (FXN), which functions as iron chaperone [172]. Friedreich's ataxia (FRDA), caused by decreased expression of FXN, is characterized by mitochondrial iron accumulation, mitochondrial dysfunction and increased oxidative stress. Of note, ferroptosis inhibitors have been effectively tested as potential therapeutic approach on primary FRDA patient-derived fibroblasts [173]. Recently, Jing Du et al have demonstrated a link between FXN and ferroptosis in cancer. Suppression of FXN impairs mitochondrial morphology, prevents Fe-S cluster assembly and enhances CDI ferroptosis in HT-1080 cancer cells [173].
Mitochondria contain a labile iron pool which is extremely redox active [25]. In physiological conditions, free iron homeostasis is tightly controlled by FtMt [174]. FtMt is structurally similar to cytosolic FtH and has ferroxidase-and iron-binding activities similar to cytosolic ferritin [175]. Mitochondrial ferritin protects against mitochondrial ROS accumulation [176] that, otherwise, may injure proteins, lipids and DNA within the mitochondria and impair ATP production, causing energy stress [177]. Downregulation of FtMt enhances mitochondrial free iron accumulation and inevitably leads to mitoROS accumulation and ferroptosis [76].
The importance of the mitochondrial iron metabolism in ferroptosis is further supported by the role of the new discovered iron-sulfur proteins (2Fe-2S) NEET. These proteins mediate the export of sulfur ions and iron between the cytosol and the mitochondria [178]. Deletion of the mitochondrial isoform CDGSH iron sulfur domain 1 (CISD1), also known as mitoNEET, causes mitochondrial iron accumulation and generation of mitochondrial lipid peroxides contributing to ferroptosis [29]. Interestingly, CISD1 knockdown mice exhibit many features of Parkinson's disease [179].

Ferroptosis Mediated by Mitochondrial VDACs
Voltage-dependent anion channels (VDAC1, VDAC2, and VDAC3) operate at the OMM to control the trafficking of ions and metabolites between cytosol and mitochondria [168]. Consequently, loss of VDAC2/3 affects mitochondrial activity by disrupting ∆Ψ m homeostasis. In this regard, Yagoda et al. [28] have demonstrated that binds to and targets VDAC2/3 resulting in ∆Ψ m alteration and eventually ferroptosis in cancer cells harboring RAS mutations. In agreement, siRNA-mediated knockdown of VDAC2/3 is able to attenuate erastin-induced ferroptosis [78]. In the same study, Yagoda et al. [28] have shown that erastin treatment also breaks down the expression of both VDAC2 and VDAC3. In melanoma cells, erastin induces the activation of the E3-ligase Nedd4 which, in turn, induces VDAC2/3 ubiquitination and ferroptosis [169]. Similarly, RSL3 treatment causes VDAC2/3 degradation through Nedd4 [26]. However, knockdown of VDAC2/3 suppresses the sensitivity of cells to erastin but not to RSL3 [180].
Although VDACs have been largely considered constitutively open, recent studies show that VDACs conductance capacity is inhibited by intracellular free tubulin abundance [181]. In preclinical models of osteosarcoma, microtubule-destabilizing agents increase cytoplasmic free tubulin causing a decrease of ∆Ψ m [182]. Erastin and other analogues block the tubulin-dependent VDAC closure, thus leading to an increase of ∆Ψ m and ROS-dependent mitochondrial dysfunction, bioenergetic failure and, ultimately, ferroptosis in HepG2 and Huh7 human hepatocarcinoma cells [36].

Other Pathways
Erastin-induced ferroptosis is linked to mitochondrial transactivation of Bcl-2 family member BH3-interacting domain death agonist (BID) (Figure 3) [27]. This protein acts as a connection bridge between surface death receptors (e.g., Fas and tumor necrosis factor-α, TNF-α) and the core extrinsic apoptotic pathway in mitochondria [183]. Its activation is mediated by caspase-8 cleavage [184]; BID is then translocated into the mitochondria where it activates the pro-apoptotic proteins BAX and BAK [185]. Of note, knockout of BID in neural cells, using CRISPR/Cas9 approach, preserves mitochondrial integrity and function, and mediates a neuroprotective effect against ferroptosis [27]. The specific mechanisms of BID in paradigm of ferroptosis need further investigations.
Lon peptidase 1 (LONP1) mediates the selective degradation of misfolded or oxidatively damaged polypeptides in the mitochondrial matrix and maintain the integrity of the mitochondrial genome ( Figure 3) [186]. In the PANC1 cell line, erastin-induced ferroptosis enhances the expression of mitochondrial LONP1. Conversely, LONP1 inhibition leads to the activation of NRF2/KEAP1 signalling pathway and to the up-regulation of GPX4, thus inhibiting ferroptosis [187].
An overview of the crucial mitochondrial actors involved in ferroptosis induction are reported in Table 4. forms an activating thioester bond between the fatty acid and CoA [34] CS catalyzes the first reaction of the TCA, condensing acetyl-CoA and oxaloacetate to form citrates [26,34] Iron metabolism markers IRON alterations in (Fe-S) clusters and LIP amount contribute to accumulation of ROS [14,29] Mfrn1/2 iron accumulation and oxidative damage [30,32,167,188] FtMt protects against the increase of mitochondrial ROS though its storage and ferroxidase activity [76] FLVCR1b iron export mechanism out of mitochondria [189] ABCB7/8 mitochondrial Fe-S cluster export [190,191] CISD1 regulates mitochondrial iron uptake and generation of mitochondrial lipid peroxides [29] Others VDAC2/3 control the trafficking of ions and metabolites between cytosol and mitochondria, leading an enhanced absorption of mitochondrial iron [169,170] FSP1 mitochondrial effector of apoptotic cell death, able to convert CoQ10 in ubiquinol, that traps lipid peroxyl radicals [57] BID acts as a connection bridge between surface death receptors and the core apoptotic pathway in mitochondria [27] LONP1 mediates the selective degradation of misfolded or oxidatively damaged polypeptides in the mitochondrial matrix and maintain the integrity of the mitochondrial genome [187] This table reports mitochondrial proteins and the relative molecular mechanisms regulating ferroptosis.

Discussion
Ferroptosis occurs when lipid hydroperoxide detoxification mediated by GPX4 activity is reduced to such an extent that it becomes insufficient to restrain iron-dependent membrane PUFA oxidation and toxic ROS accumulation [2].
As a main source of cellular ROS, mitochondrial metabolism is likely to play a pivotal role in the execution of ferroptosis [26]. A survey of the literature clearly highlights that ferroptosis is accompanied by severe morphological and functional mitochondrial damages and that, at the same time, a proper function of mitochondrial bioenergetic metabolism is mandatory for the initiation and the accomplishment of this new type of cell death [35,45,99]. Interference of key regulators of mitochondrial lipid metabolism (i.e., ASCF2 and CS), glutamine metabolism (i.e., GLS2), TCA cycle (i.e., FH) and other signaling pathways consistently enhance sensitivity to ferroptosis [26,34]. Nonetheless, our knowledge of the molecular mechanisms underlying these events are still limited and additional studies are warranted.
Suggestive evidence of the ferroptosis/mitochondria crosstalk is represented by the strong iron dependency of this RCD [26,38]. Intracellular iron accumulation can generate ROS and cause oxidative stress via Fenton Reaction, thereby promoting lipid peroxidation [61]. Mitochondrial iron homeostasis is altered to satisfy the redox active iron demands for propagating ferroptosis [26,99]. A direct in vivo evidence for the involvement of mitochondrial iron metabolism in ferroptosis is represented by neurodegenerative diseases, whose pathogenetic mechanisms have been recently linked to ferroptosis [167]. Whether mitochondrial iron crosstalk with cytosolic iron or, otherwise, mitochondrial iron metabolism is independent, to a certain extent, from cytosolic iron metabolism is still under debate [16]. For instance, when heme synthesis is inhibited in the mitochondrion, iron continues to enter these organelles [192]. This finding may suggest the lack of communication between cytoplasm and mitochondrion, as iron continues to be transported into this organelle irrespective of heme synthesis inhibition. Otherwise, it can suggest that iron continues to enter the mitochondrion in an effort to rescue heme synthesis. A recent work by Li et al. [193], highlighted that fibroblasts and lymphoblasts from Friedreich's ataxia (FA) patients display cytosolic iron-deficiency. Overexpression of mitochondrial ferritin (FtMt) in the mitochondrion leads to mitochondrial iron-loading and cytosolic iron deprivation [194]. Collectively, these data suggest that mitochondrial iron metabolism can mediate ferroptosis by modulating whole-cell iron processing.
Metabolic plasticity is a critical property that gives cancer cells the edge for expanding, persisting after therapeutic hits and evading immune surveillance [195]. Recently, metabolic reprogramming has been associated with acquired sensitivity to ferroptosis, thus opening up new opportunities to treat therapy-insensitive tumors [1]. Of note, either the genetic manipulation or the pharmacological targeting of proteins involved in ferroptosis have been found to induce cell death in a wide range of cancer cells [11]. The susceptibility of different types of cancer cells to ferroptosis is though significantly variable [155]. Based on some recent studies, the different sensitivity of cancer cells to ferroptosis depends on their basic metabolic status [78]. Considering the pivotal role of mitochondria in tumor cell metabolic rewiring, it is possible that modulation of the mitochondrial metabolic pathways might reshape the tumor microenvironment thus leading to ferroptosis-mediated tumor suppression. To make some examples, cancer stem cells frequently present a mitochondrial metabolic shift from glycolysis to OXPHOS [196], that can be exploited to make these cells vulnerable to ferroptosis. Glutaminolysis is used by the majority of cancer cells to satisfy their bioenergetic requirements [197]. Since its role in promoting ferroptosis, glutaminolysis may represent a nodal point of vulnerability for cancer cells and a potential target for novel anti-tumor strategies [198]. Iron addiction is a characteristic of cancer cells [199]. Modulation of both mitochondrial FXN and NEET proteins has been associated with CDI ferroptosis in cancer cells.
Overall, these findings provide a clear support for the potential use of mitochondria-mediated ferroptosis in cancer treatment. Future studies exploring the effects of mitochondrial metabolic rewiring in in vivo models of ferroptosis would be necessary to confirm the role of this cell death as new exciting frontier in cancer biology.