Defining a Pharmacological Inhibitor Fingerprint for Oxytosis/Ferroptosis

Abstract

Ferroptosis was first described in 2012 as an iron- and lipid peroxidation-dependent form of regulated cell death. Since its initial description, these two characteristics have informed numerous cell culture studies where inhibitors of lipid peroxidation and/or iron chelators have been shown to prevent cell death induced by a wide range of insults. However, it is not clear whether these two characteristics are sufficient to distinguish ferroptosis from other forms of regulated cell death. Thus, the primary goal of this study was to determine whether a unique combination of features could be identified that would provide an approach to more clearly separate ferroptosis from other forms of regulated cell death. To this end, multiple pharmacological inhibitors based on a variety of studies were tested. Many of these inhibitors were previously shown to protect cells from oxytosis, a regulated cell death pathway that mechanistically overlaps with ferroptosis and is induced by some of the same chemicals as ferroptosis. These inhibitors were not only tested against both known ferroptosis and oxytosis inducers but also a number of other insults that have been suggested to induce ferroptosis. The results show that a pharmacological fingerprint for ferroptosis can be established and used to categorize toxic insults into those that overlap with oxytosis/ferroptosis and those that do not.

Introduction

Ferroptosis has been described as an iron and lipid peroxidation-dependent form of regulated cell death [1]. Based on these two characteristics, numerous studies have suggested that a wide range of inducers of cell death work through ferroptosis. However, it is not at all clear that these two characteristics are sufficient to define ferroptosis, even in cell-based studies. Indeed, a recent review reported the dependency of several other forms of regulated cell death on iron and/or lipid peroxidation [2]. Thus, the primary goal of this study was to determine whether a unique combination of features could be identified that would provide a pharmacological fingerprint of ferroptosis for cell culture studies that also could be used to inform in vivo analyses. Since it was important that this fingerprint be relatively easy to implement, experiments that involve genetic knockdowns or overexpression were only used to provide validation for the different pharmacological features that were chosen for the fingerprint.

Observations on oxytosis, a form of regulated cell death that was first described over 20 years ago and shares many, if not all, of the characteristics of ferroptosis [3, 4], were taken as a starting point for inhibitor selection. Importantly, oxytosis proceeds via a stepwise pathway [5] which has been further elaborated upon over the past 20 years [4, 6]. Oxytosis is usually initiated by the inhibition of the cystine/glutamate transporter, system x_c⁻, by glutamate leading to a drop in glutathione (GSH) levels. The ferroptosis inducer erastin also inhibits system x_c⁻ [1] and cystine deprivation has an analogous effect [7]. Once the GSH levels fall significantly, ROS levels start to increase exponentially [8]. A major source of these ROS appears to be the mitochondrial electron transport chain (ETC) [8, 9]. Mitochondria initially become hyperpolarized during the course of this cell death process [5, 9–11] but then depolarize near to the time of cell death [12]. Other sources of ROS, including NADPH oxidases [13] and lysosomes [14], can also contribute to the increase in ROS during oxytosis.

GSH depletion also results in the inhibition of glutathione peroxidase 4 (GPx4) activity since this enzyme depends on an adequate supply of GSH for its function [15]. Both GPx4 inhibition and GSH depletion lead to the activation of 12/15 lipoxygenases (12/15-LOX) [15, 16] and the production of lipid hydroperoxides. GPx4 is unique in its ability to reduce lipid hydroperoxides, the substrates of 12/15-LOX, embedded in membranes [17]. Importantly, calcium influx is required for cell death [8, 18–20] and 12/15 LOX inhibitors prevent the rise in intracellular calcium following GSH depletion [16]. At about the same time in the cell death process, the pro-apoptotic Bcl-2 family member Bid (BH3-interacting domain death agonist) translocates to the mitochondria with Bid-loaded mitochondria accumulating around the nucleus and losing their membrane integrity [12]. However, markers of apoptosis, including chromatin condensation are not observed [3].

Inhibitors of each of these steps were used to identify a set of compounds that would reliably block cell death initiated by the system x_c⁻ inhibitors glutamate and erastin as well as the GPx4 inhibitor RSL3. In addition, the genetic studies on ferroptosis that have identified specific proteins that either enhance or protect from cell death were exploited to identify additional potential ferroptosis inhibitors [21–23]. The recent literature was also mined for compounds with defined targets and pathways that were described as inhibiting ferroptosis and these were also included in the initial screen [10, 24–26]. There was no deliberate exclusion of any compound or pathway. By combining these sources, 20 compounds were tested and from these, a set of 17 inhibitors that block oxytosis and ferroptosis were identified (Figure 1). This set of inhibitors was then tested against a number of other toxic insults that have been reported or speculated to induce ferroptosis [27]. Our studies clearly show that while the protective effects of one or two inhibitors might suggest that a toxic insult activates ferroptosis, this is not the case when the full pharmacological fingerprint is taken into consideration. The results also suggest that several of the hallmarks of oxytosis/ferroptosis appear to be shared with other cell death pathways suggesting that compounds that prevent these changes could have benefits beyond blocking oxytosis/ferroptosis.

Figure 1:

Oxytosis/Ferroptosis pathway showing the targets of most of the pharmacological inhibitors tested.

Finally, not only were these studies carried out in the HT22 hippocampal mouse nerve cell line, which has been the basis of many of the studies on oxytosis [4], but also in HT1080 cells, a human tumor cell line that has been used in many of the studies on ferroptosis [1, 28] and human MC65 cells, a cell culture model of intracellular amyloid toxicity [29] that we recently showed results in induction of the oxytosis/ferroptosis pathway [30], to determine if the panel was generally useful among different cell lines. Overall, these studies provide important, additional insight into the role of diverse signaling pathways in oxytosis/ferroptosis.

Materials and Methods

Materials

Bafilomycin (11038), BI-6C9 (17265), bisindoylmaleimide III (11072), chlorgyline (15925), erastin (17754), trifluoromethoxycarbonylcyanide phenylhydrazone (FCCP; 15218), (ferrostatin (17729), Flt 3 inhibitor (21193), GKT137831 (17764), idebenone (15475), liproxstatin (17730), LY83583 (70230), mitoquinol (mitoQ; 89950), Nullscript (16433), PD146176 (10010518), PI3K inhibitor VIII (10009210), RSL3 (19288), Scriptaid (10572), STY-BODIPY (27089), 2-theonyl trifluoroacetone (TTFA; 15517) and troglitazone (71750) were purchased from Cayman Chemical (Ann Arbor, MI USA). GSK2795039 (HX18950) was from MedChem Express (Monmouth Junction, NJ USA). Cisplatin (2251) was from Tocris (Minneapolis, MN USA). Antimycin A (A8674), (aminooxy)acetic acid hemihydrochloride (AOA; C13408), apomorphine (A4393), CdCl₂ (202908), 8-(4-chlorophenylthio)-guanosine 3’,5’-cyclic monophosphate (pCPT-cGMP; C5438), CoCl₂ (C2644), deferiprone (379409), glutamate (G5889), glutaminase inhibitor 968 (SML1327), hydrogen peroxide (H₂O₂; H1009), 6-hydroxydopamine (6OHDA; H4391), iodoacetic acid (I4386), myxothiazol (T5580), paraquat (36541), rotenone (45656), sodium azide (S8032), tert-butyl hydroperoxide (tBOOH; B2633) and all other chemicals were purchased from Sigma-Aldrich (St. Louis, MO USA) unless otherwise stated.

Oxytosis/Ferroptosis Assay

For this assay, 5 x 10³ HT22 mouse hippocampal nerve cells, grown in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal calf serum (FCS) (Hyclone, Logan, UT, USA), were plated per well in 96-well dishes. After 24 hr of culture, the medium was exchanged with fresh medium and the various inducers of cell death were added at the concentrations indicated in the Tables or Figure Legends with or without the inhibitors which were used at the concentrations listed in Table 1 and Figure 7. Dose range finding studies were done with the inhibitors prior to deciding on the doses reported in Table 1 and Figure 7. After 24 hr of treatment, viability was measured by the 3-(4, 5-dimethylthiazolyl-2)-2,5-diphenyltetrazolium bromide (MTT) assay as previously described [31]. Results obtained from the MTT assay correlated directly with the extent of cell death as confirmed visually. Controls with the inhibitors alone were used to test for toxicity and/or effects on cell proliferation.

Table 1:

Potential Oxytosis/Ferroptosis Inhibitors

Compound	Target	5 mM glutamate	500 nM erastin	250 nM RSL3
No treatment		7.3 ± 4.3%	6.5 ± 3.7%	7.2 ± 6.9%
Ferrostatin (10 μM)	Lipid perox.	83.1 ± 4.0%****	82.6 ± 2.7%****	84.7 ± 3.3%****
Liproxstatin (1 μM)	Lipid perox.	75.1 ± 5.7%****	71.5 ± 4.4%****	81.9 ± 1.5%****
Deferiprone (100 μM)	Fe chelator	90.0 ± 5.0%****	93.2 ± 2.7%****	87.8 ± 3.7%****
MitoQ (1 μM)	mitochondria	90.9 ± 13.0%****	76.8 ± 6.9%****	102.8 ± 13.4%****
Clorgyline (100 μM)	mitochondria	86.4 ± 8.7%****	86.0 ± 4.7%****	95.4 ± 7.1%****
968 (10 μM)	mitochondria	88.4 ± 1.4%****	70.3 ± 1.1%****	79.5 ± 8.3%****
AOA (1 mM)	mitochondria	79.0 ± 4.2%****	78.3 ± 8.3%****	1.5 ± 0.2%
GSK2795039 (10 μM)	NOX2	78.1 ± 0.5%****	71.7 ± 1.7%****	87.5 ± 1.7%****
GKT137831 (10 μM)	NOX1/4	63.7 ± 6.6%****	64.2 ± 5.0%****	90.0 ± 2.4%****
PD146176 (5 μM)	15LOX	83.9 ± 0.6%****	88.9 ± 7.5%****	87.4 ± 1.5%****
Troglitazone (1 μM)	ACSL4	76.5 ± 4.4%****	70.3 ± 4.2%****	96.8 ± 4.0%****
Idebenone (1 μM)	FSP1	70.2 ± 5.3%****	68.9 ± 4.2%****	91.7 ± 7.7%****
CoCl2 (100 μM)	Calcium influx	77.8 ± 1.1%****	72.9 ± 4.0%****	76.4 ± 6.8%****
LY83583 (1 μM)	Calcium influx	76.0 ± 2.6%****	81.1 ± 2.6%****	91.2 ± 1.3%****
Apomorphine (5 μM)	Calcium influx	86.9 ± 1.0%****	78.6 ± 7.2%****	90.7 ± 4.2%****
BI-6C9 (10 μM)	Bid inhib.	84.7 ± 2.5%****	78.7 ± 8.1%****	90.9 ± 1.0%****
Bafilomycin (100 nM)	Autophagy	88.7 ± 1.3%****	91.2 ± 7.2%****	60.3 ± 4.8%****
Scriptaid (10 μM)	HDAC	73.0 ± 10.9%****	69.7 ± 6.0%****	81.8 ± 2.1%****
Nullscript (10 μM)	HDAC	5.9 ± 7.2%	10.9 ± 8.6%	10.8 ± 12.5%
PI3K I VIII (250 nM)	PI3 kinase	82.7 ± 6.4%****	81.7 ± 5.7%****	0%
Flt3 inhibitor (1 μM)	Flt3	76.1 ± 1.0%****	69.8 ± 1.3%****	88.7 ± 1.0%****
Bisindolemal. (10 μM)	PKC	16.1 ± 4.8%	25.4 ± 5.0%***	28.4 ± 3%****

Potential oxytosis/ferroptosis inhibitors were tested for their ability to protect mouse HT22 hippocampal cells against glutamate, erastin and RSL3 toxicity at doses that induce 85%-95% cell death. Initially, a range of inhibitor concentrations were tested based on literature reports. The most effective concentrations are reported here. The values presented are the average of a minimum of five independent experiments with all treatments done in triplicate.

^***p<0.001;

^****p<0.0001 versus glutamate, erastin or RSL3 alone.

Figure 7:

Final pharmacological fingerprint and heat map of cell death inducers versus the inhibitors.

For some assays, human HT1080 cells (ATCC #CCL-121) were used. In this case, the cells were grown in the same medium as the HT22 cells but plated at 1 xl0⁴ cells/well for the toxicity assays. The cells were treated identically to the HT22 cells in all other respects except that higher concentrations of glutamate, erastin and RSL3 were needed to kill the cells.

Intracellular Amyloid Toxicity in MC65 Neural Cells

The MC65 neural cells were obtained from Dr. Bryce Sopher (University of Washington). The induction of intracellular Aβ toxicity in MC65 nerve cells was performed exactly as described [30]. MC65 nerve cells express the C99 fragment of the amyloid precursor protein under the control of a tetracycline-mediated promoter. The cells are grown in DMEM with 10% FCS and 2 μg/ml tetracycline. The removal of tetracycline induces C99 synthesis and C99 is cleaved to Aβ. By day 3, cells in the absence of tetracycline are dead, while control, uninduced cells remain viable. Cell viability was determined by the MTT assay as described above. In all cases, cells in the dishes were examined microscopically before the addition of the MTT reagent to ensure that any positive results in the MTT assay are not an artifact due to interaction of the extracts with the assay chemistry. The data are presented as viability relative to controls plus tetracycline.

ATP Assay

HT22 cells were plated at 1.5 x 10⁵ cells/dish in 35 mm dishes. After 24 hr of culture, the medium was exchanged with fresh medium and the various inducers of cell death were added at the concentrations indicated in the Figure Legends. At the indicated times, the cells were scraped into lysis buffer and assayed for ATP levels using the Invitrogen ATP determination kit (#A22066; ThermoFisher, USA) per the manufacturer’s instructions as described previously [31]. ATP levels were normalized to total protein levels in the cell extracts as determined using the bicinchoninic acid (BCA) assay (ThermoFisher).

Inhibition of Autoxidation of Egg-PC Liposomes

Egg phosphatidylcholine (EggPC) liposomes (1 mM) (#840051C; Avanti, USA) and STY-BODIPY (1.5 μM) in TBS at pH 7.4 were added to an opaque/clear bottom 96-well plate. The different test compounds were then added at a final concentration of 10 μM. The concentration of 10 μM was chosen based on the effect of ferrostatin in this assay. ≥10 μM ferrostatin protects against oxytosis/ferroptosis and also inhibits ~75% of autoxidation of 1 mM egg-PC liposomes and 1.5 μM STY-BODIPY at 4 hr, the time needed for the STY-BODIPY to be completely oxidized in the presence of DMSO under the same conditions. All the compounds were tested at the same concentration to be able to directly compare their inhibitory effects on lipid peroxidation in this cell-free system in relation to ferrostatin. The plate was vigorously mixed and incubated at 37°C for 10 min in a plate reader (SpectraMax). The plate was ejected and the autoxidation was initiated by the addition of 0.5 mM V-70 (LB-V70; Wako, Japan). The plate was mixed again and incubated at 37°C for 4 hr. Data was acquired every 15 minutes by excitation of STY-BODIPY at 488 nm and emission collected at 518 nm. The data are shown as the concentration of oxidized STY-BODIPY (ox-STY-BODIPY) determined by dividing the maximal relative fluorescent units (RFU) levels in DMSO by the concentration of the dye, resulting in a response factor of 1.59 x 10³ RFU/μM.

Protein Preparation and Western Blotting

For Western blotting, 1.5 x 10⁵ HT22 cells per 35 mm dish were grown for 24 hr prior to the indicated treatments. Total protein extracts were prepared by rinsing the cells twice with ice-cold phosphate-buffered saline. The cells were scraped into lysis buffer (50 mM HEPES, pH 7.5, 50 mM NaCl, 50 mM NaF, 10 mM Na pyrophosphate, 5 mM EDTA, 1% Triton X-100, 1 mM Na₃VO₄, 1 x protease inhibitor cocktail, 1 x phosphatase inhibitor cocktail) and incubated on ice for 30 min. Extracts were sonicated and cleared by centrifugation. The supernatants were stored at −70°C until analysis. Protein concentrations were quantified by the BCA assay and adjusted to equal concentrations. 5x Western blot sample buffer (74 mM Tris-HCl, pH 8.0, 6.25% SDS, 10% β-mercaptoethanol, 20% glycerol) was added to a final concentration of 2.5x and samples were boiled for 5 min.

For SDS-PAGE, equal amounts of cellular protein, typically 10-20 μg per lane, were used. All samples were separated using 4-12% Criterion XT Precast Bis-Tris Gels (Biorad, Hercules, CA). Proteins were transferred to nitrocellulose membranes and the quality of protein measurement, electrophoresis and transfer checked by staining with Ponceau S. Membranes were blocked with 5% non-fat milk in TBS-T (20 mM Tris buffer pH 7.5, 0.5 M NaCl, 0.1% Tween 20) for 1 hr at room temperature and incubated overnight at 4°C in the primary antibody diluted in 5% BSA in TBS/0.05% Tween 20. The primary antibodies used were: rabbit anti-ATG5 (#8540, 1/1000), rabbit anti-ferritin heavy chain-1 (#4393, 1/10,000) rabbit anti-LC3 (#4108, 1/1000) and HRP-conjugated rabbit anti-actin (#5125, 1/10,000) from Cell Signaling and rabbit anti-HIF1α (#10006421, 1/1000) from Cayman Chemical. The transfers were rinsed with TBS/0.05% Tween 20 and incubated for 1 hr at room temperature in horseradish peroxidase-goat anti-rabbit or goat anti-mouse (Biorad) diluted 1/5000 in 5% nonfat milk in TBS/0.1% Tween 20. The immunoblots were developed with the Super Signal reagent (Pierce, Rockford, IL). For all antibodies, the same membrane was re-probed for actin. Autoradiographs were scanned using a Biorad GS800 scanner. Band density was measured using the manufacturer’s software. Each Western blot was repeated at least three times with independent protein samples.

Transfection

For siRNA transfection, HT22 cells were plated in 60 mm dishes at 5 x 10⁵ cells/dish and 20 pmol ATG5 siRNA (#sc-41446 mouse; #sc41445- human) or HIF1α siRNA (#sc-35562) from Santa Cruz Biotechnology or control siRNA (#1027280) from Qiagen, were used along with RNAi max (Invitrogen) according to the manufacturer’s instructions.

Statistical Analysis

All of the experiments were done at least in triplicate and repeated at least three times. Data from at least three independent experiments were normalized, pooled and analyzed using Graph Pad Prism 9 software. The results were analyzed for statistically significant differences at the p < 0.05 level using the t-test or analysis of variance (ANOVA) test and Tukey’s post test for individual group means comparisons as appropriate.

Results

In carrying out these studies, we realized that it was critical to evaluate the ability of the different inhibitors to protect against similar levels of cell death. Thus, for all of the cell death inducers tested, including glutamate, erastin and RSL3, concentrations that induce 85-95% cell death within 24 hr were used. For glutamate, erastin and RSL3, all of the inhibitors were added at the same time as the insult but in the case of some of the other cell death inducers, a 1 hr pretreatment with the inhibitors was tested if little or no effect was seen when the inhibitors were added at the same time as the insult. Initially, multiple concentrations of all of the different inhibitors were tested against the insults in order to determine the optimal concentration for protection or, if there was no protection, then the highest dose that could be tested without causing significant toxicity.

Identification of inhibitors of different cellular processes that protect against glutamate, erastin and RLS3

The two sets of compounds frequently used to define ferroptosis are lipid peroxidation inhibitors, such as liproxstatin and ferrostatin, and iron chelators, such as deferiprone [32]. As shown in Table 1, these compounds provided excellent protection against cell death induced by glutamate, erastin or RSL3.

However, in the case of iron chelators such as deferiprone there remained the question of whether part of their protective activity is due to the induction of HIF-1α, since this transcription factor is induced by iron chelation [33] and was reported to be protective against oxytosis/ferroptosis [for review see 26]. To address this question, HIF1α expression was downregulated with siRNA (Fig. 2A) and then the effects on the ability of the iron chelator deferiprone to protect the cells against glutamate, erastin and RSL3 were examined. Loss of HIF1α did not have any effect on protection by deferiprone (Fig. 2B). Similar to iron chelators, CoCl₂ can also induce HIF1α [33] so the effects of HIF1α downregulation on protection by CoCl₂ were also examined. Similar to the results with deferiprone and the HIF1α knockdown, there was no loss of protection by the calcium channel blocker CoCl₂ (Fig. 2B) when HIF1α levels were decreased by ~95% (Fig. 2A) indicating that the protective effect of CoCl₂ in this model also is not mediated by HIF1α. Thus, these results indicate that induction of HIF1α is not part of the protective activity of iron chelators or the calcium channel blocker CoCl₂ against oxytosis/ferroptosis in HT22 cells.

Figure 2: Effect of HIF1α knockdown on protection by deferiprone and CoCl₂.

HIF1α was knocked down using specific siRNA as described in Materials and Methods (A). Cells treated with control siRNA or HIF1α siRNA were then assayed for the ability of either deferiprone (DF) or CoCl₂ (Co) to protect from glutamate (glu, 5 mM), erastin (eras, 500 nM) or RSL3 (250 nM) toxicity (B).

A number of years ago, it was shown that the ROS generated from the mitochondrial ETC played a key role in oxytosis [8, 9] with both mitochondrial ETC inhibitors and mitochondrial uncouplers providing protection [8]. A recent study that used cystine deprivation to induce oxytosis in HT1080 cells obtained similar results [11]. However, in the same study, the authors found that ETC inhibitors did not block ferroptosis induced by RSL3 in HT1080 cells. These results appeared to be in contrast to those obtained in a study which showed that in HT22 cells, as well as mouse embryo fibroblasts (MEF), RSL3 induced an increase in mitochondrial ROS, as determined using MitoSox staining, and the mitochondrial-specific antioxidant MitoQ prevented cell death [34]. Thus, it was decided to further evaluate the effects of ETC inhibitors (Table 2) as well as MitoQ (Table 1) on cell death induced by glutamate, erastin and RSL3. As shown in Table 2 and consistent with the results of Gao et al [11], ETC inhibitors effectively blocked cell death caused by treatment with both glutamate and erastin but did not reduce RSL3-induced cell death. These inhibitors include the complex I inhibitor rotenone, the complex II inhibitor TTFA, the complex III inhibitors antimycin A and myxothiazol, the complex IV inhibitor NaN₃ and the uncoupler FCCP [9]. However, consistent with the results of Jelinek et al [34], MitoQ robustly blocked cell death induced by all three insults (Table 1).

Table 2:

ETC Inhibitors

Inhibitor	glu	erastin	RSL3	IAA
No treatment	7.8 ± 0.4%	14.1 ± 4.4%	12.1 ± 2.5%	7.2 ± 1.3%
Rotenone (1 μM)	79.2 ± 11.6%****	70.7 ± 1.0%****	0%	15.3 ± 1.6 %
TTFA (100 μM)	79.1 ± 6.6%****	65.5 ± 25.3%***	0%	30.2 ± 4.0%*
Antimycin A (1 μM)	67.9 ± 10.4%****	73.2 ± 12.5%****	0%	5.7 ± 9.8%
Myxothiazol (1 μM)	75.3 ± 4.9%****	82.1 ± 2.3%****	0%	0%
NaN3 (5 mM)	89.9 ± 1.2%****	85.2 ± 4.3%****	0%	10.7 ± 9.5%
FCCP (0.5 μM)	63.3 ± 6.2%****	72.6 ± 2.6%****	0%	44.3 ± 17.7%***

ETC inhibitors were tested for their ability to protect mouse HT22 hippocampal cells against glutamate, erastin and RSL3 toxicity at doses that induce 85%-95% cell death. Initially, a range of ETC inhibitor concentrations was tested based on literature reports. The most effective concentrations are reported here. The values presented are the average of a minimum of three independent experiments with all treatments done in triplicate.

^*p<0.05;

^***p<0.001;

^****p<0.0001 versus glutamate, erastin or RSL3 alone.

Previously, it was shown that clorgyline reduced ROS production from complex 1 of the ETC [8]. Surprisingly, clorgyline also blocked cell death by all three insults (Table 1) and was particularly effective against RSL3 toxicity (Fig. 3A). Indeed, the IC₅₀ for clorgyline-mediated protection against RSL3 was ~5x lower than for protection against either glutamate or erastin. However, clorgyline is also a MAO A inhibitor and MAO A is located in the mitochondria and generates ROS [35] suggesting that MAO A activity may be a source of RSL3-induced mitochondrial ROS independently of the ETC.

Figure 3:

(A) Increased efficacy of clorgyline against RLS3 toxicity. HT22 cells in 96 well plates were treated overnight with the indicated concentrations of clorgyline and glutamate (5 mM), erastin (500 nM) or RSL3 (250 nM). Cell survival was measured the next day by the MTT assay. (B) Time dependent changes in ATP levels following treatment of HT22 cells with glutamate (5 mM), erastin (500 nM) or RSL3 (250 nM). Cells were treated in 35 mm dishes and then harvested at the indicated times for measurement of ATP levels. (C) Time dependent changes in ATP levels following treatment of HT22 cells with H₂O₂ (750 μM), t-BOOH (5 μM) or IAA (20 μM). (D) Time dependent changes in ATP levels following treatment of HT22 cells with paraquat (PQ, 2.5 mM), 6OHDA (500 μM), CdCl₂ (50 μM) or cisplatin (100 μM). Cells were treated in 35 mm dishes and then harvested at the indicated times for measurement of ATP levels.

In addition to the ETC, the TCA cycle has also been implicated in oxytosis/ferroptosis [24]. Specifically, these authors demonstrated that glutamine contributed to cell death by being converted to glutamate and subsequently α-ketoglutarate and that both the glutaminase inhibitor compound 968 and the transaminase inhibitor amino-oxyacetate (AOA) blocked cell death induced by erastin and cystine deprivation. Interestingly, while compound 968 prevented glutamate-, erastin- and RSL3-induced cell death, AOA did not inhibit RSL3-induced cell death (Table 1) suggesting that AOA may act on additional targets upstream of GPx4 in oxytosis.

Given the multiple roles of mitochondria in cell death, the effects of the oxytosis/ferroptosis inducers on energy production were explored further. As shown in Figure 3B none of the three inducers had a significant effect on ATP levels over the time period prior to the point when the cells began to die (1-6 hr for RSL3 and 4-8 hr for glutamate and erastin).

NADPH oxidases (NOXs) can also be a source of ROS and have been implicated in oxytosis [13]. Since multiple NOXs are thought to contribute to nerve cell death [36], the protective effects of GSK2795039, a NOX2 inhibitor and GKT137831, a dual NOX1 and NOX4 inhibitor that does not have antioxidant activity at concentrations up to 100 μM [37] were tested. As shown in Table 1, both GSK2795039 and GKT137831 provided very good protection against glutamate and erastin and excellent protection against RSL3.

Over 20 years ago, it was shown that LOXs are activated by GSH loss and play a key role in the execution of oxytosis [16]. This observation was extended by Seiler et al [15] who showed that GPx4 inhibition is upstream of LOX activation. More recently, a number of other studies have shown a requirement for LOX activity in oxytosis/ferroptosis and specifically, as was suggested in our original paper [16], membrane bound LOXs [38]. Thus, the 15-LOX inhibitor PD146176 was included in our testing and showed excellent protection against glutamate, erastin and RSL3 (Table 1).

Several years ago, acyl-CoA synthetase long-chain family member 4 (ACSL4) was identified as an essential component of the ferroptotic pathway by promoting lipid peroxidation [21]. Thus, both triacsin C, a general ACSL inhibitor [39], and troglitazone, which was shown to inhibit RSL3 or GPX4 knockdown-induced cell death at least partly through inhibition of ACSL4 [21], were tested. Triacsin C by itself was quite toxic to the HT22 cells and provided only variable protection (not shown). However, troglitazone effectively inhibited cell death induced by glutamate, erastin or RSL3 (Table 1), consistent with the earlier data.

Two recent studies [22, 23] demonstrated a protective role against ferroptosis for apoptosis inducing factor mitochondria-associated 2 (AIFM2) which was renamed ferroptosis suppressor protein 1 (FSP1). This protein works through ubiquinone (CoQ10) to trap lipid peroxyl radicals. Idebenone is a short chain synthetic analog of CoQ10 which was shown many years ago to protect cells against oxytosis [18] and also protects cells lacking FSP1 from RSL3-induced death [22]. The results shown in Table 1 on the protective effects of idebenone against glutamate-, erastin- and RSL3-induced cell death are consistent with these studies.

The first reports on oxytosis induced either by cystine deprivation or glutamate demonstrated a requirement for calcium influx as cell death was blocked by low calcium, the calcium chelator EDTA or the calcium channel blocker CoCl₂ [18, 19]. Importantly, calcium influx was found to be a late step in the cell death process [8, 19, 40] occurring after the loss of GSH and the production and accumulation of ROS, including lipid peroxides generated by the activation of LOXs [8, 40]. In addition, calcium influx also required the activation of soluble guanyl cyclases (sGC), a target of LOX-generated lipid hydroperoxides [40], and was antagonized by the activation of dopamine D4 receptors in HT22 cells and primary neurons [41]. Thus, the effects of the calcium uptake inhibitor CoCl₂, the sGC inhibitor LY83583 and the dopamine receptor agonist apomorphine on erastin and RSL3 toxicity with glutamate toxicity as a positive control were examined. As shown in Table 1, all three compounds were highly effective against glutamate, erastin and RSL3 toxicity strongly suggesting that calcium influx plays a key role in cell death induced by all of these insults.

Elevated cGMP levels contribute to calcium influx and the cell permeable and phosphodiesterase resistant cGMP analogue, pCPT-cGMP, can potentiate oxytosis [20, 40]. Thus, it was asked if pCPT-cGMP also potentiated cell death induced by erastin and RSL3. Two approaches were taken. In the first, fixed concentrations of glutamate, erastin or RSL3 that only kill ~15-20% of the cells were used and the ability of increasing concentrations of pCPT-cGMP to potentiate cell death were tested. As shown in Figure 4A, pCPT-cGMP dose dependently enhanced the toxicity of all three compounds to a very similar extent. In the second approach, a fixed concentration (1 mM) of pCPT-cGMP was used and the effects on the percent cell death induced by increasing concentrations of glutamate, erastin or RSL3 were examined (Fig. 4B–D). In all cases, pCPT-cGMP potentiated the toxicity of the compounds when the compounds alone killed 15% or more of the cells. Surprisingly, the strongest effects were seen with RSL3.

Figure 4: Potentiation of the toxicity of oxytosis/ferroptosis inducers by pCPT-cGMP.

(A) HT22 cells in 96 well plates were treated with vehicle (cGMP alone) or fixed concentrations of glutamate (2 mM), erastin (100 nM) or RSL3 (50 nM) and the indicated concentrations of pCPT-cGMP. Cell survival was measured the next day by the MTT assay. ***p < 0.001 relative to pCPT-cGMP alone. HT22 cells in 96 well plates were treated with (B) glutamate alone or glutamate + 1 mM pCPT-cGMP; (C) erastin alone or erastin + 1 mM pCPT-cGMP; and (D) RSL3 alone or RSL3 + 1 mM pCPT-cGMP. Cell survival was measured the next day by the MTT assay. *p <0.05; **p < 0.01; ***p < 0.001 relative to glutamate, erastin or RSL3 alone.

Over 10 years ago, the pro-cell death BH3-only Bcl2 family protein Bid (BH3-interacting domain death agonist) was shown to play a late role in oxytosis in HT22 cells [12]. Both Bid knockdown and Bid inhibition by the small molecule BI-6C9 protected cells from glutamate toxicity [12]. Cell death induced by overexpression of an activated form of Bid was also inhibited by BI-6C9. Subsequently it was found that BI-6C9 did not prevent early increases in lipid peroxidation induced by glutamate [42]. Consistent with these results, BI-6C9 provided strong protection against glutamate, erastin and RSL3 toxicity (Table 1).

Multiple studies have shown that autophagy plays an important, pro-death role in ferroptosis [43, 44] and oxytosis [45, 46]. Not surprisingly, 100 nM bafilomycin provided excellent protection against glutamate, erastin and RLS3 (Table 1). To further confirm the pro-cell death role of autophagy in oxytosis/ferroptosis, the critical autophagy protein ATG5 was downregulated with siRNA as described previously [47]. Loss of ATG5 provided significant protection against glutamate, erastin and RSL3 (Fig. 5A). Consistent with these observations, it was found that the LC3II to LC3I ratio, an indicator of autophagy activation [48], was increased by all three oxytosis/ferroptosis inducers and that the levels of ferritin heavy chain, a critical target of autophagy in this context [49], were reduced (Fig. 5C, D).

Figure 5:

(A) Effects of ATG5 knockdown on glutamate (5 mM), erastin (500 nM) or RSL3 (250 nM)-induced cell death in HT22 cells. ATG5 was knocked down with specific siRNA as described in Materials and Methods and then the sensitivity to the oxytosis/ferroptosis inducers was assayed. *p <0.05; ***p < 0.001 relative to glutamate, erastin or RSL3 alone. (B) Effects of ATG5 knockdown on glutamate (50 mM), erastin (2.5 μM) or RSL3 (1 μM)-induced cell death in HT1080 cells. ATG5 was knocked down with specific siRNA as described in Materials and Methods and then the sensitivity to the oxytosis/ferroptosis inducers was assayed. Inset shows level of knockdown. (C) Time dependent changes in the LC3II/LC3I ratio and the levels of ferritin heavy chain (FTH) following treatment of HT22 cells with glutamate or erastin. (D) Time dependent changes in the LC3II/LC3I ratio and the levels of ferritin heavy chain (FTH) following treatment of HT22 cells with RSL3. Cells were treated in 35 mm dishes and then harvested at the indicated times for Western blotting using the indicated antibodies. Inset shows representative blots. Graphed results are the average of 3 independent experiments.

Several studies have shown that histone deacetylase (HDAC) inhibitors can inhibit oxytosis/ferroptosis induced by glutamate or homocysteic acid (HCA), another system x_c⁻ inhibitor, in nerve cells [28]. Thus, we also tested both the HDAC inhibitor Scriptaid and its negative control, Nullscript. As expected, Scriptaid but not Nullscript provided excellent protection against glutamate, erastin and RSL3 (Table 1)

Finally, we examined several inhibitors that were reported to block oxytosis/ferroptosis at undefined steps in earlier studies including the PI3Kα inhibitor VIII and a Flt3 inhibitor [10] and the PKC inhibitor bisindolemaleimide [25]. These inhibitors were all tested at multiple concentrations within the ranges that were reported to be effective in the earlier studies. Bisindolemaleimide provided only very weak protection against all of the inducers of oxytosis/ferroptosis while the PI3Kα inhibitor did not prevent RSL3 toxicity (Table 1). In contrast, the Flt3 inhibitor was equally protective against all three insults (Table 1).

Testing of the inhibitor fingerprint against multiple toxicities potentially associated with ferroptosis

Based on these results it was decided to include in the preliminary pharmacological fingerprint the 17 inhibitors that strongly protected against all three toxicities as indicated in Table 1 as well as the negative HDAC inhibitor control Nullscript. For the first test of this fingerprint, the effects of these inhibitors against two frequently used inducers of cell death, hydrogen peroxide (H₂O₂) and t-butyl hydroperoxide (t-BOOH), were tested. Importantly, both compounds have been suggested to induce ferroptosis [50–52]. As shown in Table 3, there was moderate protection by several of the inhibitors against t-BOOH toxicity including ferrostatin, liproxstatin, and the Flt3 inhibitor and strong protection by deferiprone, MitoQ, bafilomycin and Scriptaid. In contrast, the only inhibitor that strongly protected against H₂O₂ was deferiprone while CoCl₂ provided moderate protection. Importantly, multiple compounds tied to inhibition of the oxytosis/ferroptosis pathway, including compound 968, GSK2795039, GKT137831, PD146176, troglitazone, idebenone, LY83583, apomorphine and BI-6C9, provided no protection against either insult. The time dependent effects of the two compounds on ATP levels were also examined. As shown in Figure 3C, while H₂O₂ caused a large, rapid decrease in ATP levels, t-BOOH had little effect until shortly before the cells began to die. Thus, these results strongly suggest that neither H₂O₂ nor t-BOOH induce ferroptosis. The data also indicate that iron chelators alone cannot be used to distinguish between oxytosis/ferroptosis and other forms of cell death since deferiprone provided excellent protection against H₂O₂ despite it clearly not inducing oxytosis/ferroptosis.

Table 3:

H₂O₂, tBOOH and IAA Toxicity

Compound	H2O2 (750 nM)	tBOOH (5 μM)	IAA (15 μM)
No treatment	4.3 ± 5.6%	5.5 ± 2.0%	4.1 ± 4.3%
Ferrostatin (10 μM)	22.4 ± 23%	58.4 ± 3.0%****	83.0 ± 4.2%****
Liproxstatin (1 μM)	19.9 ± 22%	59.6 ± 6.2%****	74.5 ± 3.4%****
Deferiprone (100 μM)	86.5 ± 23.1%****	78.4 ± 12.4%****	92.0 ± 2.5%****
MitoQ (1 μM)	0%	88.5 ± 7.1%****	5.5 ± 3.3%
Clorgyline (100 μM)	0%	27.4 ± 11.5%	14.0 ± 4.0%
968 (10 μM)	0%	11.4 ± 19.7%	80.5 ± 5.7%****
GSK2795039 (10 μM)	0%	3.4 ± 5.7%	76.4 ± 4.0%****
GKT137831 (10 μM)	1.8 ± 3.2%	2.9 ± 3.4%	74.2 ± 2.8%****
PD146176 (5 μM)	0%	4.4 ± 6.1%	79.8 ± 7.0%****
Troglitazone (1 μM)	0%	24.2 ± 4.3%	55.5 ± 0.6%****
Idebenone (1 μM)	0%	0%	78.7 ± 4.3%****
CoCl2 (100 μM)	56.6 ± 5.2%****	18.0 ± 4.1%	34.1 ± 14.3%****
LY83583 (1 μM)	0%	0%	0%
Apomorphine (5 μM)	0%	0%	79.0 ± 3.5%****
BI-6C9 (10 μM)	0%	0%	72.9 ± 2.2%****
Bafilomycin (100 nM)	25.0 ± 2.0%	74.4 ± 1.3%****	73.8 ± 3.2%****
Scriptaid (10 μM)	0%	71.4 ± 6.9%****	72.7 ± 2.4%****
Nullscript (10 μM)	0%	20.5 ± 8.0%	59.5 ± 7.5%****
Flt3 inhibitor (1 μM)	24.6 ± 2.8%	58.2 ± 8.7%****	78.2 ± 6.7%****

The oxytosis/ferroptosis inhibitors that were effective against glutamate, erastin and RSL3 as shown in Table 1 were tested for their ability to protect mouse HT22 hippocampal cells against H₂O₂, tBOOH and IAA toxicity at doses that induce 85%-95% cell death. The inhibitor concentrations that were most effective in the studies shown in Table 1 were used. The values presented are the average of a minimum of three independent experiments with all treatments done in triplicate.

^****p<0.0001 versus H₂O₂, tBOOH or IAA alone.

We then turned to a toxicity that our lab has studied for a number of years, in vitro ischemia [31]. In this assay, the irreversible inhibitor of the glycolytic enzyme glyceraldehyde 3-phosphate dehydrogenase, iodoacetic acid (IAA) [53], is used to cause a rapid loss of ATP [Figure 3C and 31]. It has been shown that this assay can identify compounds that maintain ATP levels in the presence of stress [31] and it has been used as a primary screen to identify novel and highly potent anti-oxytotic derivatives of the flavonoid fisetin [54]. Interestingly, as shown in Table 3, this cell death pathway has much greater overlap with oxytosis/ferroptosis than H₂O₂ or tBOOH toxicity. Out of the 17 inhibitors, 10 protected strongly, 2 protected moderately, 3 provided little or no protection. In addition, Scriptaid did not protect much better than the negative control, Nullscript, suggesting that its effect in this context was non-specific. The major differences were with inhibitors that either affect mitochondrial ROS production (MitoQ and clorgyline), the ETC (Table 2) or protect at late steps in the cell death process, especially with respect to calcium influx.

Given these results, it was asked whether cell death induced by several other toxins that have been suggested to cause oxytosis/ferroptosis [27, 55] would be inhibited by the battery of inhibitors. To this end, 6-hydroxydopamine (6OHDA) and paraquat (PQ), two compounds relevant to Parkinson’s disease, cadmium chloride (CdCl₂) and the anti-cancer drug cisplatin were tested. Surprisingly, most of the inhibitors had little or no effect on any of these toxicities (Table 4). The exceptions were the excellent protection provided by deferiprone against 6OHDA toxicity and MitoQ against cisplatin toxicity. In addition, CoCl₂ provided moderate protection against 6OHDA. Scriptaid also provided moderate protection against 6OHDA as well as cisplatin toxicity although in the latter case, the negative control also protected quite well. The inducers also had distinct effects on ATP levels with only PQ causing a rapid loss (Fig. 3D). These results indicate that all four of these insults induce forms of cell death that are quite distinct from oxytosis/ferroptosis. They also further highlight that protection against cell death by an iron chelator does not indicate that the cells die via oxytosis/ferroptosis.

Table 4:

6OHDA, PQ, CdCl₂, cisplatin

Compound	6OHDA (500 μM)	PQ (2.5 mM)	CdCl2 (50 μM)	Cisplatin (100 μM)
No treatment	14.2 ± 2.7%	10.6 ± 5.3%	6.7 ± 4.0%	26.7 ± 5.0%
Ferrostatin (10 μM)	18.8 ± 5.3%	19.1 ± 4.0%	11.6 ± 9.5%	31.6 ± 5.2%
Liproxstatin (1 μM)	20.0 ± 6.5%	15.0 ± 5.1%	11.7 ± 4.5%	30.5 ± 3.7%
Deferiprone (100 μM)	100 ± 5.0%****	14.5 ± 1.3%	14.9 ± 10.0%	31.4 ± 6.3%
MitoQ (1 μM)	30.6 ± 6.7%*	1.1 ± 1.6%	0%	60.7 ± 6.4%****
Clorgyline (100 μM)	22.7 ± 9.5%	4.9 ± 3.9%	6.1 ± 6.2%	31.2 ± 13%
968 (10 μM)	22.3 ± 6.5%	27.9 ± 8.5%***	2.1 ± 2.1%	38.2 ± 4.4%
GSK2795039 (10 μM)	10.5 ± 1.6%	2.2 ± 2.4%	15.2 ± 3.7%	27.7 ± 6.1%
GKT137831 (10 μM)	21.1 ± 6.4%	7.0 ± 2.7%	4.2 ± 1.4%	21.7 ± 1.2%
PD146176 (5 μM)	34.9 ± 10.3%**	33.7 ± 8.2%****	6.8 ± 4.8%	39.3 ± 6.6%
Troglitazone (1 μM)	24.4 ± 6.4%	16.1 ± 10.7%	5.8 ± 0.8%	33.6 ± 1.9%
Idebenone (1 μM)	18.5 ± 7.3%	3.4 ± 4.7%	6.0 ± 3.1%	26.3 ± 4.1%
CoCl2 (100 μM)	44.0 ± 2.1%****	4.2 ± 4.2%	0%	33.1 ± 6.7%
LY83583 (1 μM)	27.1 ± 7.9%	5.0 ± 0.8%	0%	0%****
Apomorphine (5 μM)	24.4 ± 8.8%	3.9 ± 4.3%	0.8 ± 1.1%	29.0 ± 2.4%
BI-6C9 (10 μM)	29.8 ± 5.0%	9.2 ± 1.7%	8.2 ± 4.1%	38.4 ± 5.2%
Bafilomycin (100 nM)	26.1 ± 4.1%	0.6 ± 0.06%	4.0 ± 1.2%	28.5 ± 4.7%
Scriptaid (10 μM)	51.7 ± 2.6%****	0%	5.7 ± 5.7%	53.2 ± 9.5%***
Nullscript (10 μM)	21.4 ± 5.7%	0%	4.6 ± 4.1%	41.6 ± 10%
Flt3 inhibitor (1 μM)	15.5 ± 8.6%	1.3 ± 1.8%	12.3 ± 5.9%	25.7 ± 6.6%

The oxytosis/ferroptosis inhibitors that were effective against glutamate, erastin and RSL3 as shown in Table 1 were tested for their ability to protect mouse HT22 hippocampal cells against 6OHDA, PQ, CdCl₂ and cisplatin at doses that induce 85%-95% cell death except for cisplatin where the maximum cell death never exceeded ~75%. The inhibitor concentrations that were most effective in the studies shown in Table 1 were used. The values presented are the average of a minimum of three independent experiments with all treatments done in triplicate.

^*p<0.05;

^**p<0.01;

^***p<0.001;

^****p<0.0001 versus 6OHDA, PQ, CdCl₂ or cisplatin alone.

Testing the in vitro radical trapping antioxidant activity of the oxytosis/ferroptosis inhibitors

Based on the contrasting levels of protection provided by some of the inhibitors tested here against different insults associated with oxidative stress, it was asked if their anti-oxytotic/ferroptotic activity was due to the inhibition of their specific enzymatic targets or possibly caused by their actions as non-specific radical trapping antioxidants, which could inhibit the cell death process by directly stopping the oxidation of lipids [56]. Thus, several compounds, including the Flt3 inhibitor, the 15LOX inhibitor PD146176, the NOX inhibitors GSK2795039 and GKT137831 and the sGC inhibitor LY83583 were tested in a cell-free, liposome-based assay to determine if they could directly inhibit lipid peroxidation. As shown in Figure 6, neither PD146176 nor LY83583 showed any significant anti-lipid peroxidation activity, while the NOX inhibitors both presented only a modest level of activity. In contrast, the Flt3 inhibitor displayed very potent inhibition of lipid peroxidation in this assay suggesting that the protection provided by this compound against oxytosis/ferroptosis and/or other insults in cells is likely to be, at least in part, associated with its ability to buffer lipid hydroperoxides and not with the inhibition of Flt3.

Figure 6: In vitro radical trapping antioxidant activity of the oxytosis/ferroptosis inhibitors.

Co-autoxidation of STY-BODIPY (1.5 μM) and the polyunsaturated lipids of egg-phosphatidylcholine liposomes (1 mM). The indicated compounds (10 μM) were added to the STY-BODIPY/liposome mix and incubated at 37°C for 10 min. Then, lipid autoxidation was initiated using V-70 (0.5mM) and monitored by measuring the fluorescence increase over time at 37°C. Graph shows the average of n=4 representative experiments.

Testing the inhibitor fingerprint in human fibrosarcoma HT1080 cells treated with glutamate, erastin or RLS3

To determine if the pharmacological fingerprint could be used in distinct cell types, the different inhibitors were tested against glutamate, erastin and RSL3 toxicity in human HT1080 cells, a commonly used model for studying ferroptosis. These cells are less sensitive to erastin and RSL3 and much less sensitive to glutamate as compared to the HT22 cells (Table 5). Nevertheless, concentrations of the inducers that killed 85-95% of the cells were identified and used to test the different inhibitors. As shown in Table 5, almost all of the inhibitors that protected against the oxytosis/ferroptosis inducers in HT22 cells also protected the HT1080 cells. The exceptions were bafilomycin as well as Scriptaid which had previously been shown to be ineffective in HT1080 cells [28]. Since the result with bafilomycin was unexpected, a wide range of doses were tested but none provided protection (not shown). In addition, ATG5 knockdown was also tested and proved to be equally ineffective as bafilomycin at preventing cell death in the HT1080 cells (Fig. 5B).

Table 5:

HT1080 Cell

Compound	50 mM glutamate	2.5 μM erastin	1 μM RSL3
No treatment	10.1 ± 4.7%	11.7 ± 2.5%	14.2 ± 2.8%
Ferrostatin (10 μM)	59.4 ± 1.3%****	75.7 ± 5.9%****	90.8 ± 4.8%****
Liproxstatin (1 μM)	56.9 ± 3.5%****	72.8 ± 1.9%****	88.6 ± 3.7%****
Deferiprone (100 μM)	71.0 ± 4.1%****	73.0 ± 5.0%****	83.4 ± 5.1%****
MitoQ (1 μM)	90.2 ± 7.1%****	89.8 ± 3.3%****	64.3 ± 2.3%****
Clorgyline (100 μM)	66.3 ± 12.8%****	73.3 ± 3.2%****	72.0 ± 4.5%****
968 (10 μM)	67.4 ± 8.9%****	69.8 ± 7.0%****	81.7 ± 6.7%****
GSK2795039 (10 μM)	64.8 ± 7.4%****	69.6 ± 7.0%****	67.1 ± 4.5%****
GKT137831 (10 μM)	63.7 ± 6.6%****	70.7 ± 7.8%****	76.9 ± 9.6%****
PD146176 (5 μM)	66.9 ± 7.8%****	72.2 ± 8.1%****	49.7 ± 4.6%****
Troglitazone (1 μM)	65.5 ± 5.6%****	72.0 ± 1.0%****	86.2 ± 5.1%****
Idebenone (1 μM)	78.8 ± 3.6%****	74.9 ± 2.5%****	77.5 ± 2.5%****
CoCl2 (100 μM)	61.4 ± 5.4%****	53.8 ± 2.9%****	56.1 ± 6.2%****
LY83583 (1 μM)	64.2 ± 3.8%****	68.6 ± 2.2%****	86.0 ± 6.7%****
Apomorphine (5 μM)	86.9 ± 1.0%****	54.4 ± 4.9%****	93.6 ± 7.1%****
BI-6C9 (10 μM)	60.6 ± 1.2%****	69.6 ± 1.5%****	77.9 ± 6.8%****
Bafilomycin (100 nM)	25.5 ± 4.0%	18.0 ± 6.6%	4.0 ± 2.7%
Scriptaid (10 μM)	16.9 ± 9.5%	15.6 ± 3.7%	12.0 ± 1.0%
Nullscript (10 μM)	3.8 ± 1.7%	0.2 ± 0.3%	5.0 ± 3.0%
Flt3 inhibitor (1 μM)	71.7 ± 4.2%****	70.1 ± 2.5%****	90.9 ± 8.7%****

The oxytosis/ferroptosis inhibitors that were effective against glutamate, erastin and RSL3 in the HT22 cells as shown in Table 1 were tested for their ability to protect human HT1080 cells against glutamate, erastin and RSL3 toxicity at doses that induce 85%-95% cell death. Initially, the same inhibitor concentrations that were effective in the HT22 cells were tested. If no protection was seen, then a range of concentrations was tested. The most effective concentrations are reported here. The values presented are the average of a minimum of three independent experiments with all treatments done in triplicate.

^****p<0.0001 versus glutamate, erastin or RSL3 alone.

Testing the inhibitor fingerprint in the human MC65 cell model of intracellular Aβ toxicity

Recently, we showed that intracellular amyloid beta (Aβ) accumulation in a cellular model of Alzheimer’s disease led to metabolic reprogramming and the induction of the oxytosis/ferroptosis pathway based on analysis of transcriptomic, proteomic and metabolomic data [30]. Thus, it was asked if the pharmacological fingerprint also applied to this model wherein Aβ accumulation induces oxytosis/ferroptosis. As shown in Table 6, almost all of the inhibitors provided excellent to outstanding protection against Aβ induction in the MC65 cells. The only exceptions were bafilomycin and MitoQ which provided little or no protection at any of the doses examined. It should be noted that the doses of the inhibitors that provided protection in the MC65 cells without causing significant toxicity were, in some cases, quite different from the doses that were effective in the HT22 or HT1080 cells. Thus, for all of the inhibitors, extensive dose response curves were done to determine if a protective dose could be determined in the MC65 cells.

Table 6:

MC65 Cells

Compound	-tet
No treatment	14.2 ± 2.7%
Ferrostatin (10 μM)	77.5 ± 7.2%****
Liproxstatin (1 μM)	97.2 ± 5.9%****
Deferiprone (100 μM)	95.8 ± 2.5%****
MitoQ (10 nM)	39.4 ± 5.1%*
Clorgyline (10 μM)	82.1 ± 1.1%****
968 (5 μM)	64.6 ± 1.8%****
GSK2795039 (10 μM)	98.0 ± 1.3%****
GKT137831 (10 μM)	106.8 ± 16%****
PD146176 (2.5 μM)	91.4 ± 7.1%****
Troglitazone (1 μM)	89.1 ± 5.7%****
Idebenone(1 μM)	77.1 ± 2.1%****
CoCl2 (100 μM)	88.4 ± 2.0%****
LY83583 (1 μM)	80.3 ± 2.4%****
Apomorphine (5 μM)	96.3 ± 24.5%****
BI-6C9 (5 μM)	112.8 ± 13.4%****
Bafilomycin (100 nm)	4.9 ± 6.1%
Scriptaid (250 nm)	64.1 ± 8.3%****
Nullscript (250 nm)	1.7 ± 2.7%
Flt3 inhibitor (1 μM)	101.3 ± 9.5%****

The oxytosis/ferroptosis inhibitors that were effective against glutamate, erastin and RSL3 as shown in Table 1 were tested for their ability to protect human MC65 cells against intracellular Aβ toxicity. Initially, the same inhibitor concentrations that were effective in the HT22 cells were tested. If no protection or toxicity was seen, then a range of concentrations was tested. The most effective concentrations are reported here. The values presented are the average of a minimum of three independent experiments with all treatments done in duplicate.

^*p<0.05;

^****p<0.0001 versus -tet alone.

Discussion

By testing a variety of inhibitors that act at distinct steps in the oxytosis/ferroptosis regulated cell death pathway thereby maintaining cell survival, we developed a pharmacological fingerprint that provides a much clearer definition of this form of cell death. This pharmacological fingerprint was also used to survey a number of different cell death inducers, several of which had been reported to induce ferroptosis, to determine if this was indeed the case. One of the most important discoveries that came out of this analysis is that protection by iron chelators is not specific to ferroptosis. Cell death induced by several other compounds, including H₂O₂ and 6OHDA, is potently inhibited by the iron chelator deferiprone but not by almost all of the other inhibitors tested indicating that the form of cell death induced by these compounds is distinct from ferroptosis. Similarly, the radical trapping antioxidants ferrostatin and liproxstatin protect against a cell death inducer, tBOOH, that otherwise has very little inhibitor overlap with oxytosis/ferroptosis suggesting that lipid peroxidation is also not fully specific to oxytosis/ferroptosis. Given the results of this additional testing, a simpler pharmacological fingerprint based on the inhibitors that showed a clear ability to both highlight the defining characteristics of oxytosis/ferroptosis as well as to discriminate between oxytosis/ferroptosis and other forms of cell death was developed (Figure 7). This pharmacological fingerprint should be particularly useful for cell culture studies but may also provide new approaches to studying oxytosis/ferroptosis in vivo.

In carrying out this analysis, we realized that it was critical that a dose of each cell death inducer that killed 85-95% of the cells was used in order to achieve a similar level of cell death with all of the inducers. This was done both by initial dose response studies with each cell death inducer and by always testing multiple concentrations of the inducers with the inhibitors because the percent of cell death with all inducers can vary somewhat from experiment to experiment. 85-95% cell death was chosen because with this level of cell death, any protective effect is very clear. Killing only 50% of the cells and increasing survival to 65-70% does not provide sufficient clarity because there is frequently some inter-experimental variation in the percentage of both cell survival and cell death. On the other hand, with many, if not all, of the inducers, using doses that kill 100% of the cells often results in no rescue by any inhibitor tested.

One concern that was raised regarding LOX inhibitors, and could apply to other compounds included in the fingerprint as well, is that due to the nature of their molecular structure, they act as radical trapping antioxidants rather than LOX inhibitors [56]. While this is a legitimate concern, it does not appear to be the case for most of the inhibitors about which this concern might be raised since they did not protect against tBOOH and/or IAA toxicity, two inducers where the known radical trapping antioxidants ferrostatin and liproxstatin did protect. The only significant exception to this argument is the Flt3 inhibitor. This compound provided the same level of protection as ferrostatin and liproxstatin against all of the cell death inducers tested here. Thus, to better address the question of radical trapping antioxidant activity directly, we tested the ability of the Flt3 inhibitor, as well as several of the other inhibitors, to directly block lipid peroxidation in a cell-free liposome-based assay. Consistent with the protective effects of the Flt3 inhibitor against the different cell death inducers, we found that it is a very effective inhibitor of lipid peroxidation in the cell-free system (Fig. 6). Although the inhibition of lipid peroxidation by this compound in cells was shown in a previous study, it was interpreted as a downstream effect of its actions on Flt3 activity [10]. Thus, our observation underscores the need to characterize possible off-target radical trapping activities of enzymatic inhibitors before relating certain pathways with the oxytosis/ferroptosis cell death process. Interestingly, in contrast, the 15LOX inhibitor PD146176 showed no direct inhibition of lipid peroxidation in the cell-free system (Fig. 6) and also did not provide any protection against tBOOH toxicity indicating that its protective effect is most likely due to 15LOX inhibition rather than non-specific radical trapping antioxidant activity. While the two NOX inhibitors showed modest direct inhibition of lipid peroxidation, their failure to protect against tBOOH toxicity suggests that this activity does not play a significant role in their protective effects in cells.

The mitochondrial ETC has long been known to play a role in oxytosis/ferroptosis induced by GSH depletion through its generation of ROS [57]. The ETC inhibitor results presented here strongly support that conclusion. In contrast, a recent study showed that ferroptosis induced by RSL3 does not involve the ETC and further suggested that mitochondria play no role in cell death induced by direct GPx4 inhibition [11]. However, our results with the MAO A inhibitor clorgyline and the mitochondrially-targeted antioxidant MitoQ along with the previously reported increase in mitochondrial ROS induced by RSL3 [34] suggest that there may be a role for other mitochondrial sources of ROS in cell death downstream of GPx4 inhibition. Indeed, MAO A is one of a number of enzymes that are present in mitochondria and generate ROS independently of the ETC [4]. Supporting the idea that other mitochondrial enzymes contribute to mitochondrial ROS generation, especially in the presence of RSL3, are the results with the NOX1/4 inhibitor GKT137831. NOX4 is located in mitochondria and, similar to chlorgyline, GKT137831 is also much more inhibitory against RSL3 than erastin or glutamate.

Related to the role of mitochondria is the effect of the different cell death inducers on ATP levels. In our study, cells treated with glutamate, erastin or RSL3 did not show a drop in ATP levels until just prior to cell death. Although this is not a universal observation [34], our results may be related to careful monitoring of the cells during the cell death process. Similar results were obtained with tBOOH, PQ, CdCl₂ and cisplatin. In contrast, IAA, H₂O₂ and 6OHDA caused a very rapid drop in ATP levels, well before the cells began to die. Thus, while the oxytosis/ferroptosis inducers all have a similar effect on ATP levels so do several other cell death inducers which appear to kill cells via distinct mechanisms suggesting that effects on ATP levels cannot be used to characterize this cell death pathway.

Calcium influx was shown to be an important and late part of the oxytosis pathway [8, 19, 40]. Consistent with this observation, compounds that act to reduce calcium influx, including LY83583, apomorphine and CoCl₂ were effective at protecting against glutamate as well as erastin and RSL3 toxicity. Curiously, while neither apomorphine nor LY83583 protected against H₂O₂ toxicity, CoCl₂ did provide significant protection suggesting that calcium influx mediated via a different pathway might play a role in this toxicity. Similarly, CoCl₂ also provided significant protection against 6OHDA toxicity. In contrast, none of these compounds protected against tBOOH, PQ or CdCl₂ toxicity. Furthermore, while apomorphine showed significant protection against IAA toxicity, CoCl₂ showed only modest protection and LY83583 showed no protection suggesting a distinct role for apopmorphine in protecting against this toxicity. Importantly, LY83583 is not able to inhibit lipid peroxidation in vitro, suggesting that the protection against glutamate, erastin and RSL3 provided by this compound is specifically caused by the reduction of calcium influx (Fig. 6). Thus, the inhibitor results suggest that calcium influx plays a specific role in oxytosis/ferroptosis that is distinct from the role played in other cell death pathways.

The autophagy inhibitor bafilomycin provides excellent to outstanding protection against glutamate, erastin and RSL3 toxicity in the HT22 cells. Similar results were obtained with knockdown of ATG5, a protein essential for autophagy [48]. It has been argued that inhibition of autophagy prevents the loss of the iron binding protein ferritin in a process known as ferritinophagy and this is why inhibition of autophagy is protective [49]. Interestingly, in our study no protection by bafilomcyin was seen in the HT1080 cells. This is in contrast to some reports [28, 43, 44] but not others [1] and is likely related to both different levels of baseline cell death and the time point after toxin treatment at which cell death was measured. At the high levels of cell death used in our study, which was measured after 24 hr of treatment, no protection was seen despite the fact that bafilomycin did protect from similar levels of cell death in the HT22 cells. Moreover, this is not specific to bafilomycin since ATG5 knockdown also did not protect from the oxytosis/ferroptosis inducers in the HT1080 cells.

Similar to the results with the HT1080 cells, bafilomycin also did not protect from intracellular Aβ toxicity in the MC65 cells. This is perhaps not surprising since autophagy contributes to the clearance of Aβ in these cells [58]. However, all of the other inhibitors protected the MC65 cells except for MitoQ. Since this compound was surprisingly toxic in the MC65 cells, it is quite likely that it was not possible to achieve an effective concentration thereby leading to its lack of efficacy. Nevertheless, the results with the rest of the inhibitors strongly support the conclusion that intracellular Aβ accumulation induces oxytosis/ferroptosis, underscoring the potential role of this process in Alzheimer’s disease.

This study has several limitations. First, although the pharmacological fingerprint was tested in multiple cell types relevant to oxytosis/ferroptosis, it is not clear at this time how it will perform in other cell lines derived from different tissues. This question will be addressed in future studies. Second, although we suggest a range of concentrations to test for each of the inhibitors, it could be that in some cell types a given compound is sufficiently toxic so that it is not possible to achieve an efficacious concentration. Indeed, as noted above, this possibility could explain the inability of MitoQ to protect the MC65 cells from intracellular Aβ toxicity. Third, although we tried to select inhibitors of each of the cellular processes that have been identified as relevant to oxytosis/ferroptosis, there could be additional processes that turn out to be relevant to this cell death pathway. However, this potential problem is reduced by the broad nature of the proposed pharmacological fingerprint.

In summary, this study describes a pharmacological fingerprint to define oxytosis/ferroptosis in cell culture. The use of this fingerprint clearly establishes that several widely used inducers of cell death kill cells through mechanisms distinct from oxytosis/ferroptosis despite previously published suggestions to the contrary (Fig. 7). Thus, by providing greater clarity on the mechanisms underlying cell death induced by various insults, this study should significantly help advance the field both in terms of cell culture studies as well as by suggesting potential markers for in vivo analysis.

Identification of a set of inhibitors that protect from multiple, well characterized inducers of oxytosis/ferroptosis.

Demonstration that most of these inhibitors do not protect from other inducers of oxidative stress-mediated cell death.

Protection from cell death by iron chelators and radical trapping antioxidants is not specific to oxytosis/ferroptosis.

Inhibitor fingerprint is applicable to several distinct cell types and a non-canonical inducer of oxytosis/ferroptosis.