Mitochondrial event as an ultimate step in ferroptosis

In ferroptosis, the roles of mitochondria have been controversial. To explore the role of mitochondrial events in ferroptosis, we employed mitochondrial DNA-depleted ρ0 cells that are resistant to cell death due to enhanced expression of antioxidant enzymes. Expression of mitochondrial-type GPx4 (mGPx4) but no other forms of GPx4 was increased in SK-Hep1 ρ0 cells. Likely due to high mGPx4 expression, SK-Hep1 ρ0 cells were resistant to ferroptosis by erastin inhibiting xCT channel. In contrast, SK-Hep1 ρ0 cells were susceptible to cell death by a high concentration of RSL3 imposing ferroptosis by GPx4 inhibition. Accumulation of cellular ROS and oxidized lipids was observed in erastin- or RSL3-treated SK-Hep1 ρ+ cells but not in erastin-treated SK-Hep1 ρ0 cells. Mitochondrial ROS and mitochondrial peroxidized lipids accumulated in SK-Hep1 ρ+ cells not only by RSL3 but also by erastin acting on xCT on the plasma membrane. Mitochondrial ROS quenching inhibited SK-Hep1 ρ+ cell death by erastin or a high dose of RSL3, suggesting a critical role of mitochondrial ROS in ferroptosis. Ferroptosis by erastin or RSL3 was inhibited by a more than 20-fold lower concentration of MitoQ, a mitochondrial ROS quencher, compared to DecylQ, a non-targeting counterpart. Ferroptosis of SK-Hep1 ρ+ cells by erastin or RSL3 was markedly inhibited by a VDAC inhibitor, accompanied by significantly reduced accumulation of mitochondria ROS, total peroxidized lipids, and mitochondrial peroxidized lipids, strongly supporting the role of mitochondrial events in ferroptotic death and that of VDAC in mitochondrial steps of ferroptosis induced by erastin or RSL3. SK-Hep1 ρ+ cell ferroptosis by sorafenib was also suppressed by mitochondrial ROS quenchers, accompanied by abrogation of sorafenib-induced mitochondrial ROS and mitochondrial peroxidized lipid accumulation. These results suggest that SK-Hep1 ρ0 cells are resistant to ferroptosis due to upregulation of mGPx4 expression and mitochondrial events could be the ultimate step in determining final cell fate.


INTRODUCTION
Ferroptosis is characterized by extensive accumulation of peroxidized lipids in iron-dependent manners, which is blocked by iron chelators or lipophilic antioxidants [1]. Ferroptosis is important in the progression of acute kidney injury, acetaminophen-induced liver injury, cardiac ischemia/reperfusion injury, or p53-mediated cell death [2][3][4].
However, recent investigation demonstrated important roles of mitochondria in ferroptosis [10,11]. We also have reported a pivotal role of mitochondrial lipid peroxidation and mGPx4 in doxorubicin-induced ferroptosis of cardiomyocytes [12]. Here, we studied the role of mitochondrial function and mGPx4 in ferroptosis employing mitochondrial DNA (mtDNA)-depleted SK-Hep1 ρ 0 cells that show enhanced expression of antioxidant enzymes such as mitochondrial superoxide dismutase, GPx1, or GPx4 [13,14].
We observed that SK-Hep1 ρ 0 cells are resistant to ferroptosis due to mGPx4 induction and obtained evidence suggesting that mitochondrial events are ultimate steps in ferroptosis.

RESULTS
Resistance of SK-Hep1 ρ 0 cells against ferroptosis When we studied GPx4 expression, GPx4 mRNA and protein were notably increased in SK-Hep1 ρ 0 cells compared to parental SK-Hep1 (ρ + ) cells, consistent with our previous data [14], (Fig. 1A, B). Among GPx4 isoforms, mitochondrial GPx4 (mGPx4) expression was markedly increased in SK-Hep1 ρ 0 cells compared to SK-Hep1 ρ + cells, while cytosolic GPx4 (cGPx4) or nuclear GPx4 (nGPx4) expression was not different between the two cell types (Fig. 1C). In contrast, fractionation study showed that GPx4 protein expression was enhanced in both cytosolic and mitochondrial fractions, suggesting that enhanced mGPx4 might lead to increased mGPx4 protein not only in mitochondrial fraction but also in cytoplasmic fraction (Fig. 1D), consistent with our previous data [12].
To study the mechanism of GPx4 induction in SK-Hep1 ρ 0 cells, we studied Nrf2, a master regulator of antioxidant responses and a transcriptional regulator of GPx4 [18,19]. Nrf2 expression was Fig. 1 Enhanced GPx4 expression in mtDNA-depleted SK-Hep1 ρ 0 cells and resistance to cell death by ferroptosis inducers. A mRNA expression of GPx4 in GSK-Hep1 ρ + cells and SK-Hep1 ρ 0 cells was examined by real-time RT-PCR (left) and RT-PCR (right). B GPx4 protein expression in SK-Hep1 ρ + cells and SK-Hep1 ρ 0 cells were determined by immunoblot analysis using the indicated antibodies. C Expression of different isoforms of GPx4 in SK-Hep1 ρ + cells and ρ 0 cells were studied by real-time RT-PCR. D GPx4 protein levels in the cytoplasmic and mitochondrial fractions were determined using immunoblot analysis. Purity of the cytoplasmic and mitochondrial fractions was confirmed by immunoblot analysis using anti-α-tubulin and -COXIV antibodies, respectively. E SK-Hep1 ρ + cells and ρ 0 cells were treated with 2.5 and 5 µM erastin or 0.1 and 1.0 µM RSL3 for 24 h. Cell death was measured using LDH release assay. F SK-Hep1 ρ + cells and ρ 0 cells were treated with 5 μM erastin or 0.1 μM RSL3 for 24 h in the presence or absence of 2.5 µM Fer-1 (Ferrostatin-1), 0.1 µM Lip-1 (Liproxstatin-1), 10 µM ZIL (Zileuton), 50 µM DFO (Deferoxamine), 50 µM Trolox or 50 µM z-VAD (z-VAD-FMK). Cell death was determined using LDH release assay. G Nrf2 protein level in SK-Hep1 ρ + cells and ρ 0 cells was determined using immunoblot analysis. H GPx4 protein level in SK-Hep1 ρ 0 cell extract was determined using immunoblot analysis with or without treatment with 10 µM ML385, an Nrf2 inhibitor, for 24 h. I SK-Hep1 ρ + and ρ 0 cell death after treatment with 5 µM erastin for 24 h was determined using LDH release assay with or without 10 µM ML385 or 2.5 µM Fer-1 pretreatment for 1 h. Data represent means ± SD from three independent experiments. Data were analyzed by two-tailed Student's t test (A, C, E, ρ + and ρ 0 cell comparisons in F) or one-way ANOVA with Tukey's multiple comparison test (F except ρ + and ρ 0 cell comparisons, I). (CON, control) (*p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant) (Uncropped immunoblots can be found in Fig. S2).

ROS accumulation in ferroptosis
We next studied intracellular ROS, a critical component of ferroptosis. In erastin-treated SK-Hep1 ρ + cells, massive ROS accumulation was observed by flow cytometry after staining with CM-H2DCFDA, a general ROS sensor ( Fig. 2A). In the same cells, massive peroxidized lipid accumulation was also observed by flow cytometry after labeling with BODIPY-C11, a peroxidized lipid sensor ( Fig. 2A). ROS or peroxidized lipid accumulation in erastintreated SK-Hep1 ρ + cells was markedly reduced by Fer-1 or Lip-1 ( Fig. 2A), suggesting that ROS or peroxidized lipid accumulation is causally related to erastin-induced ferroptosis. In contrast, accumulation of ROS or peroxidized lipids was absent in erastintreated SK-Hep1 ρ 0 cells ( Fig. 2A), likely due to high mGPx4 expression scavenging peroxidized phospholipids.

Mitochondrial events in ferroptosis
We next studied mitochondrial ROS which are important in several cell death models [20][21][22], since major site of mGPx4 action in SK-Hep1 ρ 0 cells might be mitochondria, despite mGPx4 protein expression in both cytoplasm and mitochondria [12]. In erastintreated SK-Hep1 ρ + cells, significant accumulation of mitochondrial ROS stained with MitoSOX, a mitochondrial ROS probe, was found ( Fig. 3A), which could be due to mitochondrial events occurring secondary to the cytosolic events during ferroptosis by erastin primarily affecting xCT on the plasma membrane or Data represent means ± SD from three independent experiments. Data were analyzed by one-way ANOVA with Tukey's multiple comparison test (A-C except ρ + and ρ 0 comparison) or two-tailed unpaired Student's t test (ρ + and ρ 0 cell comparisons in A-C). (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant).

Effects of mitochondrial ROS quenching
We further investigated the role of mitochondrial ROS in ferroptosis by employing mitochondrial ROS scavengers. When SK-Hep1 ρ + cells were pretreated with MitoTEMPO, a mitochondrial ROS quencher, cell death by erastin or 0.1 μM RSL3 was abrogated (Fig. 4A), indicating critical roles of mitochondrial ROS in ferroptosis even in that caused by erastin primarily imposing cytosolic event, while direct effect of erastin on mitochondria cannot be eliminated [24]. SK-Hep1 ρ + cell or ρ 0 cell death by 1.0 μM RSL3 was also significantly inhibited by MitoTEMPO (Fig.  4B), demonstrating critical roles of mitochondrial ROS or mitochondrial GPx4 inhibition in RSL3-induced SK-Hep1 cell ferroptosis. When we employed MitoQ, a mitochondria-targeted lipophilic ROS quencher [26], SK-Hep1 ρ + cell death by erastin or 0.1 μM RSL3 was again significantly inhibited (Fig. 4A), corroborating the role of mitochondrial ROS in ferroptosis. Consistently, inhibition of erastin-or 0.1 μM RSL3-induced SK-Hep1 ρ + cell death by MitoTEMPO or MitoQ was accompanied by suppression of mitochondrial ROS and mitochondrial lipid oxidation ( Fig.  4C-F).
To study the role of mitochondrial ROS more specifically, we compared anti-ferroptotic effects of MitoQ and DecylQ, a nontargeting counterpart [7]. DecylQ inhibited erastin-and RSL3mediated ferroptosis at more than 100-fold and 20-fold higher concentrations compared to MitoQ, respectively (IC 50 : 8.6 × 10 −7 vs 5.0 × 10−9 M for erastin and 1.6 × 10 −6 vs 7.0 × 10 −8 M for RSL3) (Fig. 4G, H), suggesting more specific and robust effects of mitochondria-targeted antioxidants compared to non-targeted antioxidants and critical roles of mitochondrial events in ferroptosis, in contrast to a previous paper [7]. The difference between inhibition of eratin-or RSL3-induced cell death by DecylQ and that by MitoQ was most conspicuous between 100 nM~1.0 μM (Fig. 4G, H). Furthermore, in this concentration range, MitoQ-mediated reduction of relative accumulation of MitoPerOx by RSL3 was more pronounced compared to that of BODIPY-C11 (Fig. S1A), suggesting that stronger effect of MitoQ on cell death compared to DecylQ is likely due to its preferential effect on mitochondrial peroxidized lipids. However, MitoQmediated reduction of erastin-induced accumulation of MitoPerOx was not different from that of BODIPY-C11 (Fig. S1B), suggesting distinct mechanism and kinetics of cytosol-mitochondrial interaction between erastin and RSL3.
To further investigate the mechanism of a dominant role of mitochondrial events in ferroptosis, we studied VDAC which can be a link between cytosolic and mitochondrial events [23,24]. When we employed an antagonist of VDAC, VBIT-4 suppressing VDAC oligomerization and permeabilization [27] almost completely abrogated erastin-induced SK-Hep1 cell death (Fig. 5A), suggesting a critical role of VDAC in erastin-induced ferroptosis. Erastin-induced accumulation of mitochondrial ROS, total peroxidized lipids and mitochondrial peroxidized lipids stained by MitoSOX, BODIPY-C11, and MitoPerOx, respectively, was also markedly suppressed by VBIT-4 ( Fig. 5B-D), suggesting roles of VDAC in mitochondrial steps of erastin-induced ferroptosis probably as a linker between cytosolic and mitochondrial events or a direct target of erastin [23,24]. We also studied the effect of VBIT-4 on RSL3-induced ferroptosis. Here again, RSL3-induced SK-Hep1 cell death was abrogated by VBIT-4 (Fig. 5A). Furthermore, RSL3-induced accumulation of mitochondrial ROS, total peroxidized lipids, and mitochondrial peroxidized lipids stained were markedly suppressed by VBIT-4 ( Fig. 5E-G), suggesting roles of VDAC in mitochondrial events in RSL3-mediated ferroptosis as well.
To prove the dominant role of mitochondrial events in ferroptosis, we infected SK-Hep1 ρ + cells with adenovirus expressing cGPx4 (Ad-cytoGPx4-FLAG) or mGPx4 (Ad-mitoGPx4-FLAG) [12]. Adenoviral mGPx4 expression almost completely abrogated SK-Hep1 ρ + cell death by erastin or 0.1 μM RSL3 (Fig.  6A), suggesting critical roles of mGPx4 and mitochondrial events of erastin-or RSL3-induced ferroptosis. Accumulation of ROS, peroxidized lipids, or mitochondrial ROS by erastin or RSL3 was also abrogated by adenoviral mGPx4 expression (Fig. 6B, C). When we infected cells with Ad-cytoGPx4-FLAG expressing cGPx4, SK-Hep1 ρ + cell death and accumulation of ROS, peroxidized lipids or mitochondrial ROS by erastin or RSL3 were also significantly inhibited ( Fig. 6A-C), which could be due to inhibition of cytosolic events of ferroptosis inducing dominant mitochondrial event, although significant roles of cytosolic event itself in ferroptosis cannot be disregarded.
We also studied whether mitochondrial events are important in ferroptosis by sorafenib inhibiting xCT [29,30], similar to erastin. Again similar to erastin, marked mitochondrial ROS and mitochondrial peroxidized lipid accumulation were observed in sorafenibtreated SK-Hep1 cells (Fig. 7E, F). Mitochondrial ROS or mitochondrial peroxidized lipid accumulation in sorafenib-treated SK-Hep1 cells was significantly reduced by Fer-1 (Fig. 7E, F), supporting roles of mitochondrial events in sorafenib-induced ferroptosis. Furthermore, MitoTEMPO or MitoQ significantly inhibited mitochondrial ROS or mitochondrial peroxidized lipid accumulation by sorafenib (Fig. 7G, H), which was accompanied by suppression of sorafenib-induced SK-Hep1 cell death (Fig. 7I), again suggesting roles of mitochondrial events in sorafenib- Fig. 4 Effect of mitochondrial ROS quenchers on ferroptosis. A SK-Hep1 ρ + cells were treated with 5 μM erastin or 0.1 μM RSL3 for 24 h in the presence or absence of 100 μM MitoTEMPO or 500 nM MitoQ. Cell death was determined by LDH release assay. B SK-Hep1 ρ + and ρ 0 cells were treated with 1.0 μM RSL3 in the presence or absence of MitoTEMPO for 24 h. Cell death was determined by LDH release assay. C, D In SK-Hep1 ρ + treated with 5 μM erastin in the presence or absence of 100 μM MitoTEMPO or 500 nM MitoQ for 16 h, accumulation of mitochondrial ROS (C) and mitochondrial lipid peroxides (D) was determined by flow cytometry after staining with MitoSOX and MitoPerOx, respectively (right). Representative histograms are shown (left). E, F In SK-Hep1 ρ + treated with 0.1 μM RSL3 in the presence or absence of 100 μM MitoTEMPO or 500 nM MitoQ for 4 h, accumulation of mitochondrial ROS (E) and mitochondrial lipid peroxides (F) was determined by flow cytometry after staining with MitoSOX and MitoPerOx, respectively (right). Representative histograms are shown (left). G, H SK-Hep1 ρ + cells were treated with 5 μM erastin (G) or 0.1 μM RSL3 (H) for 24 h in the presence or absence of varying concentrations of DecylQ and MitoQ. Cell death was determined by LDH release assay. Data represent means ± SD from three independent experiments. Data were analyzed by oneway ANOVA with Tukey's multiple comparison test (A-F) or two-tailed unpaired Student's t test (G, H). (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant).  Representative histograms are shown (upper and left lower panels). Data represent means ± SD from three independent experiments. Data were analyzed by one-way ANOVA with Tukey's multiple comparison test. (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant).
induced ferroptosis. Furthermore, DecylQ inhibited sorafenibmediated cell death at more than 10-fold higher concentration compared to MitoQ (IC 50 : 1.5 × 10 −7 vs 1.3 × 10 −8 M) (Fig. 7J), which indicates more specific and robust effects of mitochondria-targeted antioxidant compared to non-targeted antioxidant in sorafenib-induced ferroptosis caused by inhibition of xCT on the plasma membrane and suggests ultimate roles of mitochondrial events in ferroptosis.

DISCUSSION
mtDNA-depleted ρ 0 cells [31] are valuable tools to study mitochondrial functions, ROS, cell death or inflammation [14,32,33]. Since GPx4 is a critical component in ferroptosis and GPx4 expression was increased in SK-Hep1 ρ 0 cells [14], we investigated whether SK-Hep1 ρ 0 cells were resistant to ferroptosis and also studied the role of mitochondria in ferroptosis.
We observed that SK-Hep1 ρ 0 cells were resistant to cell death by erastin or a low concentration of RSL3, accompanied by a markedly reduced ROS or peroxidized lipids, which is different from previous data showing no difference in ferroptosis between parental and ρ 0 cells [34,35]. These discrepancies could be ascribed to differences in cell types, methods of ρ 0 cell generation, or the level of antioxidant gene expression in ρ 0 cells. SK-Hep1 ρ 0 cell resistance to ferroptosis together with no ROS or peroxidized lipid accumulation by erastin primarily inducing cytosolic event of ferroptosis could be due to mGPx4 scavenging peroxidized lipids not only in mitochondria but also in the cytoplasm [12]. mGPx4 induction could increase cytosolic GPx4 during transportation into mitochondria [12]. In addition to GPx4, a high expression of GPx1 detoxifying peroxidized lipids other than phospholipids in SK-Hep1 ρ 0 cells [36,37] might contribute to the reduction of ROS and peroxidized lipids.
Besides cytoplasmic action of mGPx4, mGPx4 might act on mitochondrial events as well because mitochondrial ROS or mitochondrial peroxidized lipid accumulated in erastin-treated SK-Hep1 ρ + cells and mitochondrial ROS quenchers inhibited erastin-induced SK-Hep1 ρ + cell death together with abrogated mitochondrial ROS or mitochondrial peroxidized lipid accumulation. Mitochondrial events by erastin could be due to direct action on mitochondria such as VDAC opening and changed mitochondrial outer membrane permeability [23,24], or secondary to the cytoplasmic event of ferroptosis inducing mitochondrial change such as BID transactivation [38].
In contrast to erastin inhibiting xCT on the plasma membrane, RSL3 induces ferroptosis through GPx4 inhibition and intracellular peroxidized phospholipid generation [15]. Mitochondrial ROS and mitochondrial peroxidized lipid accumulation in RSL3-treated SK-Hep1 ρ + cells could be due to the occurrence of cytosolic events of ferroptosis associated with RSL3-mediated cytosolic GPx4 inhibition leading to secondary mitochondrial changes through cellular events such as BID transactivation [38], similar to the erastin treatment. Instead, direct inhibition of mitochondrial GPx4 by RSL3 might lead to similar consequences of mitochondrial ROS and mitochondrial peroxidized lipid accumulation. Protection of RSL3-induced cell death by mitochondrial ROS quenchers accompanied by abrogation of mitochondrial ROS or mitochondrial peroxidized lipid accumulation indicates that mitochondrial events are ultimate steps in ferroptotic death, whether mitochondrial ROS was produced by secondary mitochondrial changes following RSL3-mediated cytosolic GPx4 inhibition or by direct mitochondrial event of RSL3-mitochondrial GPx4 inhibition.
We observed that VBIT-4, a VDAC antagonist, significantly inhibited SK-Hep1 cell death by erastin or RSL3, accompanied by markedly reduced accumulation of mitochondrial ROS, total peroxidized lipids, and mitochondrial peroxidized lipids, strongly supporting a critical role of mitochondrial events in ferroptosis. These data are in line with previous reports showing erastin effect on VDAC [23,24] or blockade of glutamate-induced oxytosis or acetaminophen-induced hepatic ferroptosis by a VDAC inhibitor [39,40]. However, these data are dissimilar to previous papers showing no role of VDAC3 in RSL3-induced ferroptosis [41]. In contrast, degradation of VDAC2/3 and carbonylation of VDAC2 by RSL3 have been reported [42,43], suggesting the potential role of VDAC in RSL3-mediated ferroptosis depending on cell types or context. VDAC might play a role in the transport of iron from the cytoplasm to mitochondria [44], and thereby contribute to mitochondrial steps of ferroptosis regardless of inducers.
Inhibition of erastin-induced ferroptosis by adenoviral mGpx4 expression also suggests that mitochondrial ROS could be critical, similar to the inhibition of erastin-induced ferroptotic cell death by MitoTEMPO or MitoQ. Previous papers have reported exclusive mGPx4 localization in the mitochondrial fraction [45]. However, since we observed that GPx4 from adenoviral mGPx4 expression was found in the cytoplasm during transportation of mitochondrial GPx4 [12], our adenoviral mGPx4 expression experiment alone does not confirm the sole role of mitochondrial event in ferroptosis. Nonetheless, another data showing the significant ferroptosis inhibition by mitochondrial ROS quenchers alone supports that mitochondrial events are critical steps in ferroptosis.
Furthermore, more efficient inhibition of erastin-or RSL3induced ferroptosis by MitoQ compared to DecylQ suggests that mitochondrial events could be dominant players in ferroptosis, whether mitochondrial ROS was produced by secondary mitochondrial changes following cytosolic events of ferroptosis (e.g., xCT inhibition followed by BID activation or cytosolic GPx4 inhibition) or by direct mitochondrial events (e.g., mitochondrial GPx4 inhibition or VDAC opening) (Fig. 8). Our data suggesting that mitochondrial ROS and mitochondrial lipid peroxidation are the key events in ferroptosis is in contrast to previous papers reporting that extra-mitochondrial lipid peroxidation is more important than mitochondrial lipid peroxidation in ferroptosis [7] or no accumulation of mitochondrial ROS by erastin [34]. Our results are also different from previous data showing a more efficient blockade of ferroptosis by DecylQ compared to MitoQ, suggesting a minor role of mitochondrial peroxidation in MEF ferroptosis [7]. Such differences could be due to discrepancies in the treatment method, types of ROS inducers, or species of phospholipids studied. Different cells might have distinct susceptibility to mitochondrial ROS or mitochondrial lipid peroxidation Fig. 7 Resistance of SK-Hep1 ρ 0 cells against sorafenib-induced ferroptosis. A SK-Hep1 ρ + and ρ 0 cells were treated with 10 μM sorafenib for the indicated periods, and cell death was determined using LDH release assay. B SK-Hep1 ρ + and ρ 0 cells were treated with 10 μM sorafenib in the presence or absence of 2.5 µM Fer-1, 0.1 µM Lip-1, 10 µM ZIL, 50 µM DFO, 50 µM Trolox or 50 µM z-VAD for 72 h. Cell death was determined using an LDH release assay. C, D In SK-Hep1 ρ + and ρ 0 cells treated with 10 μM sorafenib in the presence or absence of Fer-1 for 48 h, accumulation of cellular ROS (C) and lipid peroxides (D) was analyzed by flow cytometry after staining with CM-H2DCFDA and BODIPY-C11, respectively (right). Representative histograms are shown (left). E, F In SK-Hep1 ρ + and ρ 0 cells treated with 10 μM sorafenib in the presence or absence of 2.5 μM Fer-1 for 48 h, accumulation of mitochondrial ROS (E) and mitochondrial lipid peroxides (F) was determined by flow cytometry after staining with MitoSOX and MitoPerOx, respectively (right). Representative histograms are shown (left). G, H In SK-Hep1 ρ + and ρ 0 cells treated with 10 μM sorafenib in the presence or absence of 100 μM MitoTEMPO or 500 nM MitoQ for 48 h, accumulation of mitochondrial ROS (G) and mitochondrial lipid peroxides (H) was determined by flow cytometry after staining with MitoSOX and MitoPerOx, respectively (right). Representative histograms are shown (left). I After treatment of SK-Hep1 ρ + cells with 10 μM sorafenib in the presence or absence of 100 μM MitoTEMPO or 500 nM MitoQ, cell death was evaluated using LDH release assay. J After treatment of SK-Hep1 ρ + cells with 10 μM sorafenib in the presence or absence of varying doses of DecylQ or MitoQ, cell death was evaluated using LDH release assay. Data represent means ± SD from three independent experiments. Data were analyzed by one-way ANOVA with Tukey's multiple comparison test (A-I except ρ + and ρ 0 cell comparisons) or two-tailed unpaired Student's t test (ρ + and ρ 0 cell comparisons in A, J). (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant).
depending on cellular characteristics including expression of DHODH modulating mitochondrial ubiquinol [11].
We observed SK-Hep1 ρ 0 cell resistance to sorafenib which is widely used as a chemotherapeutic agent against hepatoma [28]. Although papers refuting ferroptosis in sorafenib-induced cancer cell death have been published [46], our data showing inhibition of sorafenib-induce SK-Hep1 cell death by Fer-1 or Lip-1 together with significantly reduced lipid peroxidation strongly supports sorafenib-induced ferroptosis. Furthermore, suppression of sorafenib-induced cell death by MitoTEMPO or MitoQ accompanied by reduced mitochondrial ROS or mitochondrial peroxidized lipids suggests that mitochondrial events play crucial roles in ferroptosis induced by xCT inhibitors such as erastin or sorafenib [29,30,47].
Altogether, our data suggest that mitochondrial events such as mitochondrial ROS or mitochondrial peroxidized lipid accumulation following cytosolic events by ferroptosis inducers through cytosol-mitochondria interaction or those due to direct mitochondrial effects of ferroptosis inducers play ultimate roles in ferroptotic death (Fig. 8), which could be important in cancer cell death by anti-cancer chemotherapy. Avoidance of such mitochondrial events could be a mechanism of cancer cells with aberrant or deficient mitochondrial function to resist ferroptosis by chemotherapeutic agents.

Cell death assay
Cell death was assessed using LDH release assay (G1780, Promega TM Corporation, Madison, WI, USA). Briefly, cells seeded 1 × 10 4 well in a 96well plate were exposed to various concentrations of cytotoxic compounds and inhibitors for 24 h. The culture supernatant was collected and analyzed according to the manufacturer's protocol.

RT-PCR and real-time RT-PCR
RNA was isolated using TRIzol reagent (Invitrogen, Waltham, MA, USA). cDNA synthesis was performed using 1 μg of total RNA and M-MLV Reverse Transcriptase (M1701, Promega). RT-PCR was performed using PCR PreMix (K-2016, BIONEER, Korea)and specific primers [14] for 35 cycles. Quantitative RT-PCR was performed using SYBR Green Master Mix (RR420A, Takara Bio, Kusatsu, Shiga, Japan) and specific primers in a QuantStudio 3 Real-Time PCR machine (Applied Biosystems, Waltham, MA, USA). Gene expression was normalized to that of RPL32 mRNA using the 2 −ΔCt method. cGPx4 mRNA level was determined by subtracting the 2 −ΔCt values of mGPx4 and nGPx4 from that of total GPx4.

Adenoviral infection
Ad-cytoGPx4-FLAG and Ad-mitoGPx4-FLAG recombinant adenoviral vectors were transfected into HEK293AD cells using Lipofectamine 2000 (11668027, Thermo Fisher Scientific, Waltham, MA, USA) for 7~10 days. Cells were then collected and subjected to three cycles of freezing at −80°C and thawing at 37°C. After vortexing, recombinant adenovirus containing cytoGPx4 or mitoGPx4 genes was harvested from the supernatant by centrifugation. SK-Hep1 cells were infected with Ad-cytoGPx4-FLAG or Ad-mitoGPx4-FLAG adenovirus at an MOI of 20 for 24 h for further experiments.

Cell fractionation
Cells were fractionated into mitochondria and cytoplasm using a mitochondrial/cytoplasmic fractionation kit (ab65320, Abcam), according to the manufacturer's instructions. Briefly, cells were homogenized by a pestle homogenizer, and then were centrifuged at 700 × g for 10 min. The supernatant was re-centrifuged at 10,000 × g for 30 min to collect supernatant (cytoplasmic fraction) and pellet (mitochondrion fraction). The purity of cytosolic and mitochondrial fractions was confirmed by immunoblot analysis using anti-α-tubulin and -COXIV antibodies, respectively.

Statistical analysis
Data were analyzed using the two-tailed Student's t test or one-way analysis of variance followed by Tukey's multiple comparison test. Data are presented as means ± SD of three independent experiments. p values <0.05 were considered significant.

DATA AVAILABILITY
The data generated or analyzed during the current study and materials are available from the corresponding author without imposed restriction on reasonable request.