Profiling the chemical nature of anti-oxytotic/ferroptotic compounds with phenotypic screening

Abstract

Because old age is the greatest risk factor for Alzheimer’s disease (AD), it is critical to target the pathological events that link aging to AD in order to develop an efficient treatment that acts upon the primary causes of the disease. One such event might be the activation of oxytosis/ferroptosis, a unique cell death mechanism characterized by mitochondrial dysfunction and lethal lipid peroxidation. Here, a comprehensive library of > 900 natural compounds was screened for protection against oxytosis/ferroptosis in nerve cells with the goal of better understanding the chemical nature of inhibitors of oxytosis/ferroptosis. Although the compounds tested spanned structurally diverse chemical classes from animal, microbial, plant and synthetic origins, a small set of very potent anti-oxytotic/ferroptotic compounds was identified that was highly enriched in plant quinones. The ability of these compounds to protect against oxytosis/ferroptosis strongly correlated with their ability to protect against in vitro ischemia and intracellular amyloid-beta toxicity in nerve cells, indicating that aspects of oxytosis/ferroptosis also underly other toxicities that are relevant to AD. Importantly, the anti-oxytotic/ferroptotic character of the quinone compounds relied on their capacity to target and directly prevent lipid peroxidation in a manner that required the reducing activity of cellular redox enzymes, such as NAD(P)H:quinone oxidoreductase 1 (NQO1) and ferroptosis suppressor protein 1 (FSP1). Because some of the compounds increased the production of total reactive oxygen species while decreasing lipid peroxidation, it appears that the pro-oxidant character of a compound can coexist with an inhibitory effect on lipid peroxidation and, consequently, still prevent oxytosis/ferroptosis. These findings have significant implications for the understanding of oxytosis/ferroptosis and open new approaches to the development of future neurotherapies.

Graphical Abstract

1. Introduction

There are no drugs for Alzheimer’s disease (AD) that prevent or slow down its progression. This may be due to a lack of therapies that target the primary drivers of the disease. Given that old age is the greatest risk factor for AD [1], our labs have been focused on identifying the pathological processes associated with aging that are detrimental to brain function and may be driving AD. Our efforts have led to the identification of a unique mechanism of neurodegeneration called oxytosis/ferroptosis [2, 3].

Oxytosis/ferroptosis is a form of non-apoptotic regulated cell death characterized by glutathione (GSH) depletion and dysregulated production of reactive oxygen species (ROS) from mitochondria that results in lethal lipid peroxidation [2–5]. All of these changes are detected in the aging brain and, to a larger extent, in AD [4]. Importantly, oxytosis/ferroptosis can manifest itself over a lengthy time period before the cells die, thereby offering a window for therapeutic intervention. We have shown that inhibitors of oxytosis/ferroptosis are protective in transgenic mouse models of AD and prevent the development of dementia in the SAMP8 mouse model of accelerated aging [6–9].

At the molecular level, oxytosis/ferroptosis can be triggered by glutamate which inhibits cystine uptake via system x_c⁻ and subsequently depletes intracellular GSH [2, 4, 5]. This leads to inhibition of the GSH-dependent enzyme GSH peroxidase 4 (GPX4) and activation of lipoxygenases (LOXs) [10]. GPX4 can also be directly inhibited by the chemical RSL3. Based on these two insults and using a nerve cell line-based assay to screen compounds for their protection against oxytosis/ferroptosis, we have successfully identified several drug candidates for AD [3, 11].

Here, we used this assay to screen a commercial library of > 900 natural compounds. We identified a set of potent inhibitors of oxytosis/ferroptosis, most of which were previously unreported as anti-oxytotic/ferroptotic molecules. Importantly, our data indicate that oxytosis/ferroptosis may be a common pathway in other toxicities relevant to dementia and provide clues about the chemistry of anti-oxytotic/ferroptotic compounds.

2. Materials and methods

All reagents were obtained from Sigma-Aldrich (St. Louis, MO, United States), unless otherwise stated. The library of natural compounds (cat. No. HY-L021) was purchased from MedChemExpress (NJ, United States).

2.1. Cell culture

HT22 mouse hippocampal nerve cells and MC65 human neuroblastoma cells were cultured in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen, Carlsbad, CA, United States) supplemented with 10% fetal calf serum (FCS) (Hyclone, Logan, UT, United States), and incubated at 37°C in an atmosphere with 10% CO₂.

2.2. Phenotypic screening assays

2.2.1. Oxytosis/ferroptosis (HT22 cells)

5×10³ HT22 mouse hippocampal nerve cells were plated per well in 96 well plates. After 24 h, the medium was exchanged with fresh medium and 10 mM glutamate or 200 nM RSL3 were added alone or in combination with the natural compounds at the indicated concentrations. 24 h later, the cellular viability was measured by the 3-(4, 5-dimethylthiazolyl-2)- 2,5-diphenyltetrazolium bromide (MTT) assay, as previously described [12]. Cell survival was confirmed by visual inspection of the wells. After the primary screening, compounds yielding ≥ 65% cell survival against at least one of the insults were selected to be tested at a wider range of concentrations (58 compounds), and those compounds displaying ≥ 75% protection against both insults in this secondary screening were selected for further analysis (20 compounds).

2.2.2. Intracellular Aβ toxicity (MC65 cells)

MC65 cells were regularly grown in high glucose DMEM supplemented with 10% FCS and 2 μg/ml tetracycline [13]. For the assay, cells were dissociated, plated at 4×10⁵ cells per 35 mm tissue culture dish and grown for 24 h. The next day, the cells were washed with PBS and placed in Opti-minimal essential media (Opti-MEM, Invitrogen) in the presence (no induction) or absence (APP-C99 induced) of 2 μg/ml tetracycline in combination with the natural compounds. At day 3, the control cells in the absence of tetracycline were dead, and cell viability was determined by the MTT assay and confirmed by visual inspection.

2.2.3. Protection against energy loss (HT22 cells)

HT22 cells were seeded onto 96 well plates as described in the oxytosis/ferroptosis assay. The medium was exchanged 24 h later with fresh medium and the cells were treated with 20 μM IAA alone (which results in 90–95% cell death) or in combination with the compounds at the indicated concentrations, as previously described [14]. After 2 h, the medium was replaced with fresh medium without IAA but containing the compounds. 24 h later, the cellular viability was measured by the MTT assay.

2.3. Measurement of total glutathione (tGSH)

For measurement of tGSH, 3×10⁵ HT22 cells were plated in 60 mm dishes. After 24 h, the medium was exchanged with fresh medium, and glutamate and compounds were added. The cells were treated for 24 h and then scraped into ice-cold PBS, and 10% sulfosalicylic acid was added at a final concentration of 3.3%. GSH was determined by the recycling assay based on the reduction of 5,5-dithiobis(2-nitrobenzoic acid) with glutathione reductase and NADPH [15] and normalized to protein recovered from the acid-precipitated pellet by treatment with 0.2 N NaOH at 37 °C overnight and measured by the bicinchoninic acid assay (Pierce, Rockford, IL, USA).

2.4. Measurement of ROS and lipid peroxidation levels by flow cytometry

7.5×10³ HT22 cells per well were seeded in 96 well plates. After 24 h, the compounds were added in the presence or absence of glutamate (10 mM). The concentration of each compound was chosen based on the lowest dose yielding maximal protection. After 6 h, the media was aspirated and 100 μl per well of BODIPY^™ 581/591 C11 (cat No. D3861, Invitrogen) (1μM) or CellROX^™ Green (cat No. C10444, Invitrogen) (1μM) were added in the presence of the corresponding treatments, to detect lipid peroxidation or total ROS levels, respectively. The cells were incubated for 30 minutes, washed and trypsinized. The fluorescence was measured at 4°C using a MACSQuant^® VYB Flow Cytometer (Miltenyi Biotec, Germany). The data were normalized to the emission of control cells treated with DMSO.

2.5. Fe²⁺-binding capacity

Ferrozine is a specific reagent which forms a magenta-colored complex (absorption maximum at 562 nm) with ferrous (Fe²⁺) ions [16]. Briefly, 5 μl of a ferrous chloride (FeCl₂) solution were mixed with 5 μl of compound in a 96-well plate at a 1:2 molar ratio and the mixture was incubated for 2 minutes at room temperature. 100 μl of 50 mM HEPES (pH 6.8) and 50 μl of 5mM ferrozine were then added to the mixture (50 μM final compound concentration). Absorbance (A₅₆₂) was measured immediately after addition of ferrozine using a plate spectrophotometer. Concentration of free Fe²⁺ ions corresponds linearly to the A₅₆₂. For normalization, the A₅₆₂ yielded by an equimolar solution of the standard iron-chelator deferiprone or by a control solution without compound were considered as maximum and minimum iron binding capacity, respectively.

2.6. Autoxidation of egg-phosphatidylcholine (PC) liposomes

STY-BODIPY (1.5 μM) (cat No. 27089, Cayman Chemical Co, MI, United States) and liposomes of egg-PC (1 mM) (cat No. 840051P, Avanti Polar Lipids Inc, AL, United States) in TBS (pH 7.4) were added to an opaque 96-well polypropylene plate, followed by the addition of the indicated compounds (10 μM). The autoxidation was initiated by the addition of V-70 (0.5 mM) (cat No. 001-70078, Fujifilm Wako, Japan). For the enzymatic experiments, recombinant NQO1 (cat No. MBS201141, MyBioSource, CA, United States) (50 nM), recombinant FSP1 (cat No. 29611-50, Cayman Chemical) (50 nM), NADH (75 μM), the indicated quinones (5 μM), or a combination thereof, were added to the wells containing STY-BODIPY and liposomes prior to the addition of V-70. A lower compound concentration was used here to avoid rapid consumption of NADH by the most reactive quinones. Data were acquired by excitation of STY-BODIPY at 488 nm and emission was measured at 518 nm every 15 min at 37°C. Data were transformed into [ox-STY-BODIPY] by taking the raw fluorescence values of the saturated curve of control DMSO and dividing them by the initial concentration of reduced STY-BODIPY (1.5 μM). To obtain a unique number related to the data, the area under the curve (AUC) was calculated using Graph Pad Prism 6.

2.7. Enzymatic assays

In a 96 well plate, recombinant NQO1 or FSP1 (50 nM) were mixed with 75 μM NADH in TBS buffer, pH 7.4. The reactions were initiated by adding the different quinones at 5 μM. The plate was mixed and the change in absorbance at 340 nm was measured every 15 min at 37°C. Results are expressed as concentration of NADH over time.

2.8. Statistical analysis

The half maximal effective concentration (EC₅₀) was determined from sigmoidal dose response curves using GraphPad Prism 6. All experiments were done at least in triplicate and repeated at least three times. Multiple groups were compared using one-way ANOVA followed by Dunnett’s correction. GraphPad Prism 6 was used for the statistical analyses. Data are expressed as mean ± SEM, and significance of difference is indicated as *P<0.05, **P<0.01 and ***P<0.001.

3. Results

3.1. Primary screening identifies novel potent anti-oxytotic/ferroptotic compounds

In order to identify novel anti-oxytotic/ferroptotic compounds, we screened a commercial library (HY-L021, MedChemExpress) of 903 chemically diverse compounds from a variety of biological sources including animals, microorganisms and plants (Table S1). The rationale for using this diversity was to not only increase the chances of identifying potent compounds but also to provide insight into the chemical nature of protection against oxytosis/ferroptosis.

All compounds were screened for their ability to protect HT22 cells from oxytosis/ferroptosis induced by RSL3 and glutamate, and a number of them were found to be protective at the concentrations tested (1 μM and 10 μM; Table S1). 58 compounds with ≥ 65% protection (at either 1 μM or 10 μM) against at least one of the insults were identified and re-tested at a wider range of concentrations (0.1 μM, 0.5 μM, 1 μM and 5 μM) (Table S2). This cutoff represents about two thirds of cell survival, which is a reliable level of protection that allows a good selection of compounds with good probability of displaying potent anti-oxytotic/ferroptotic activity.

In order to better estimate the anti-oxytotic/ferroptotic potency and proceed to a more detailed analysis, the range of concentrations tested was widened and the number of anti-oxytotic/ferroptotic compounds narrowed down to a manageable set of 20 compounds that protected ≥ 75% against both glutamate and RSL3 (Table 1 and Fig. 1A). The half maximal effective concentrations (EC₅₀) of these 20 compounds against RSL3 and glutamate showed a strong correlation between the protection against both insults (Table 1 and Fig. 1B).

Table 1. EC₅₀s (nM) of the top 20 anti-oxytotic/ferroptotic compounds for protection against insults relevant to AD.

RSL3 and glutamate were used to induce oxytosis/ferroptosis in HT22 cells. IAA was used to induce in vitro ischemia in HT22 cells. Aβ toxicity was induced in MC65 cells by removal of tetracycline.

Compound	EC50 vs RSL3 (nM)	EC50 vs glutamate (nM)	EC50 vs IAA (nM)	EC50 vs Aβ (nM)
β-lapachone	3 ± 1	15 ± 8	5 ± 1	5 ± 1
DHIT I	9 ± 5	59 ± 6	62 ± 18	10 ± 5
DHT I	38 ± 3	113 ± 20	26 ± 15	12 ± 7
Triptophenolide	52 ± 18	2330 ± 589	562 ± 105	40 ± 19
Tanshinone I	122 ± 35	361 ± 32	291 ± 70	100 ± 59
Idebenone	176 ± 33	182 ± 65	464 ± 187	651 ± 334
Cryptotanshinone	221 ± 14	738 ± 39	586 ± 108	19 ± 20
Lipoic acid	270 ± 80	1501 ± 131	2727 ± 465	181 ± 103
Icaritin	485 ± 87	663 ± 130	1330 ± 95	850 ± 651
Gossypol	579 ± 199	1596 ± 83	1729 ± 547	30 ± 37
L-Ascorbic acid	614 ± 82	742 ± 72	1333 ± 111	-
Baicalein	660 ± 158	1240 ± 76	1844 ± 182	253 ± 34
Catharanthine	906 ± 164	2418 ± 299	849 ± 19	2398 ± 1305
Honokiol	911 ± 211	3650 ± 148	2803 ± 19	279 ± 30
Carnosol	933 ± 104	1423 ± 88	645 ± 242	609 ± 42
3,3’ -Diindolylmethane	1086 ± 187	2714 ± 520	1798 ± 316	402 ± 39
Menaquinone-4	1317 ± 41	1173 ± 56	2747 ± 946	80 ± 65
Dopamine	1494 ± 360	1686 ± 217	4247 ± 60	3010 ± 944
Cynaroside	1152 ± 179	2439 ± 101	3856 ± 359	1203 ± 729
Fisetin	3844 ± 242	3771 ± 179	3430 ± 23	2501 ± 1173

Figure 1. Top 20 compounds and other quinones in the library.

(A) Chemical structures of the top 20 compounds. (B) Correlation between the protection against RSL3 (200 nM) and glutamate (10 mM) for the 20 most protective compounds (logEC₅₀s, nM). (C) Chemical structures of other quinones in the library.

3.2. Source and chemical profiling of the most protective compounds show an enrichment in plant-derived secondary metabolites

To investigate the relationship between the biological source of a compound and its anti-oxytotic/ferroptotic activity, we compared the sources of the top 20 protective compounds with the sources of the compounds from the entire library (Fig. 2A). The percentage of plant-derived compounds greatly increased from ~56% in the whole library to ~85% in the top 20 protective compounds. This observation suggests that plants may be a better source of anti-oxytotic/ferroptotic compounds than other sources.

Figure 2. Source and chemical classification of the anti-oxytotic/ferroptotic compounds.

(A) Pie charts showing the percentage of compounds obtained from the indicated sources. (B) Pie charts showing the percentage of plant-derived compounds belonging to the indicated chemical classes.

We then focused on the different classes of plant compounds. With the exception of quinones, the other chemical classes, including flavonoids, alkaloids, terpenoids and lignans, were represented among the most protective plant-derived compounds at similar percentages to those in the whole library (Fig. 2B). Quinone-related compounds were highly enriched in the top 20, going from ~3% of the whole library to ~35% of the top 20 protective compounds. Although four of these quinones (tanshinone I, cryptotanshinone, dihydrotanshinone I (DHT I) and dihydroisotanshinone I (DHIT I)) are derived from the same plant (Salvia miltiorrhiza) and are structurally similar (Fig. 1A), three other quinones were identified among the best 20 compounds (β-lapachone, idebenone and menaquinone-4 (vitamin K2)). This observation suggests that the class of quinones may present increased potential for compounds with anti-oxytotic/ferroptotic activity. In fact, all the quinones in the library, including those not present in the top 20 (with the exception of rhein and chrysophanol), displayed some degree of protection against oxytosis/ferroptosis (Table S3 and Fig. 1C).

Physicochemical parameters are useful for predicting the potential of a molecule to be delivered to different tissues in an organism, thus providing a preliminary estimate of how suitable a compound might be as a drug candidate. Due to the constraints of the blood-brain barrier, this type of analysis is especially relevant for drugs targeting the central nervous system (CNS) [17, 18]. Notably, most of the top 20 compounds of the library have excellent estimated physicochemical properties that are within the range of successful CNS drugs (Table 2).

Table 2. Physicochemical properties of the top 20 protective compounds.

Mw, molecular weight; ClogP, general lipophilicity; tPSA, topological polar surface area; HBD, hydrogen bond donor; HBA, hydrogen bond receptor; ClogD, lipophilicity of ionizable compounds; pKa, acid dissociation constant; MPO, multiparameter optimization score; CNS, central nervous system.

Compound	Mw	ClogP	tPSA (Å2)	HBD (n.OH,NH)	HBA (n.O,N)	ClogD (pH 7.4)	pKa (most basic center)	Pfizer CNS MPO score
Successful CNS drugs	≤ 360	≤ 5	≤ 90	≤ 3	≤ 7	≤ 2	≤ 8	0 to 6
β-lapachone	242	1.71	43	0	3	2.77	0.0	5.6
DHIT I	278	4.40	43	0	3	4.22	0.0	4.3
DHT I	278	2.43	43	0	3	3.52	0.0	5.2
Triptophenolide	312	3.64	47	1	3	4.25	9.8	3.6
Tanshinone I	276	4.80	43	0	3	4.24	0.0	4.1
Idebenone	338	3.42	73	1	5	3.41	16.0	3.9
Cryptotanshinone	296	3.37	43	0	3	4.14	0.0	4.8
Lipoic acid	206	2.39	88	1	2	−0.13	4.9	5.8
Icaritin	368	4.65	96	3	6	3.12	8.1	3.5
Gossypol	519	5.76	156	6	8	5.16	8.7	0.7
L-Ascorbic acid	176	−1.76	107	4	6	−4.99	4.2	4.4
Baicalein	270	3.00	87	3	5	1.60	6.7	5.2
Catharanthine	336	3.64	45	1	4	3.76	6.9	4.6
Honokiol	266	4.49	40	2	2	4.13	9.1	3.2
Carnosol	330	3.16	67	2	4	3.84	7.6	4.5
3,3’ -Diindolylmethane	246	4.19	32	2	2	4.83	0.0	3.5
Menaquinone-4	445	10.56	34	0	2	9.55	0.0	3.1
Dopamine	153	0.17	66	4	3	−2.18	8.6	4.7
Cynaroside	448	0.81	186	7	11	−1.21	8.0	3.4
Fisetin	286	1.24	107	4	6	1.38	7.6	4.4

3.3. Anti-oxytotic/ferroptotic activity correlates with protection against other insults relevant to AD

In order to further characterize the therapeutic potential of the top 20 protective compounds, these were tested in two additional cell culture models relevant to AD, one based on intracellular amyloid-beta (Aβ) toxicity and the other based on energy depletion.

A significant body of evidence indicates that Aβ accumulates within neurons of AD patients well before the appearance of the Aβ plaques themselves, and that this accumulation may play a central role in driving the disease [19–22]. To determine the effects of the top 20 anti-oxytotic/ferroptotic compounds on intracellular Aβ toxicity we used the MC65 nerve cell model. The MC65 is a human nerve cell line that expresses the C99 fragment of the amyloid precursor protein (APP) under the control of a tetracycline-sensitive promoter [23]. When tetracycline is withdrawn, cells express C99 which is then converted to Aβ by γ-secretase and the cells die within several days due to Aβ accumulation and aggregation within cells [24]. We found that the protection by the top 20 compounds against Aβ toxicity correlated with their protection against oxytosis/ferroptosis in HT22 cells (Table 1 and Fig. 3A).

Figure 3. Protection against oxytosis/ferroptosis correlates with protection against additional insults relevant to AD.

Correlation between protection against glutamate/RSL3 in HT22 cells and protection against Aβ toxicity in MC65 cells (A) or in vitro ischemia in HT22 cells (B) for the top 20 protective compounds.

Energy metabolism in the brain decreases with age and to larger extent in AD [25]. A breakdown in neuronal energy production leading to decreases in the levels of adenosine triphosphate (ATP) is associated with nerve cell damage and death in AD [26]. The loss of ATP can be mimicked using an in vitro ischemia model [27]. To induce ischemia, HT22 cells were treated with iodoacetic acid (IAA), a well-known irreversible inhibitor of the glycolytic enzyme glyceraldehyde 3-phosphate dehydrogenase (GAPDH) [28], and the effects of the top 20 anti-oxytotic/ferroptotic compounds in this model were evaluated. Notably, similar to the Aβ toxicity in MC65 cells, all of the compounds prevented IAA-induced cell death in a manner that highly correlated with their protection against oxytosis/ferroptosis (Table 1 and Fig. 3B). Taken together, these data are consistent with recent studies showing that the oxytosis/ferroptosis cell death pathway may underlie, at least in part, both the Aβ- and IAA-induced toxicities [13, 14, 29].

3.4. Inhibition of lipid peroxidation, but not modulation of GSH levels or total ROS production, is a common feature of the best anti-oxytotic/ferroptotic quinones

Since the group of the most potent anti-oxytotic/ferroptotic compounds was enriched in quinones, we hypothesized that these could share a protective mechanism of action. To begin investigating this hypothesis, we analyzed the effects of the top 20 compounds on the main steps of oxytosis/ferroptosis including the levels of GSH, ROS and lipid peroxidation.

To address changes in GSH metabolism, the levels of GSH were measured upon treatment with each compound in the presence or absence of glutamate (Fig. 4A). The concentration used for each compound in these assays was chosen based on the minimal concentration yielding the maximal protection against oxytosis/ferroptosis. Since we observed a strong correlation between the protection against glutamate and RSL3, the subsequent experiments were carried out only with glutamate. It was found that, although a few compounds maintained GSH levels in HT22 cells treated with glutamate (carnosol, cynaroside and fisetin), and some increased basal GSH levels in the absence of glutamate (DHIT I, DHT I, lipoic acid, carnosol, cynaroside and fisetin), most had little to no effect.

Figure 4. Modulation of GSH, total ROS and lipid peroxidation in HT22 cells exposed to oxytosis/ferroptosis.

Total GSH (A), total ROS (B) and lipid peroxidation (C) levels in HT22 cells treated with the top 20 protective compounds in the presence or absence of glutamate (10 mM). Graphs show the percentage of total GSH, fluorescent oxidized C11-BODIPY 581/591 (lipid peroxidation), fluorescent oxidized CellROX (total ROS) normalized to control DMSO treatment. The concentration of each compound tested in these assays was chosen based on the lowest dose yielding maximal protection against oxytosis/ferroptosis in HT22 cells. ^§§§p < 0.001 relative to control DMSO. *p < 0.05; **p < 0.01; ***p < 0.001 relative to control DMSO. ^#p < 0.05; ^##p < 0.01; ^###p < 0.001 relative to glutamate DMSO.

Next, the ability of the 20 best anti-oxytotic/ferroptotic compounds to modulate ROS production was determined by assessing the levels of total ROS in HT22 cells treated with the compounds in the presence or absence of glutamate. CellRox was used to measure intracellular levels of total ROS by flow cytometry. We found that about half of the compounds did not prevent the increase in ROS induced by glutamate (Fig. 4B). In fact, most of the quinones tested actually increased the levels of ROS on their own in the absence of glutamate.

Given the lack of a clear protective mechanism common to the quinones associated with regulation of GSH or total ROS levels, we then measured the effects of the compounds on lipid peroxidation. To do that, C11 BODIPY 581/591 was used to determine the levels of lipid peroxidation by flow cytometry. As seen in Fig. 4C, all of the top 20 compounds showed a strong prevention of the increase in lipid peroxidation induced by glutamate. This effect was most noticeable with the quinones, which also significantly decreased the basal levels of lipid peroxidation in the absence of glutamate.

Taken together, these data indicate that the anti-oxytotic/ferroptotic activity of the top 20 compounds analyzed, especially the quinones, is unrelated to the modulation of either ROS or GSH levels, but rather is likely dependent on their strong anti-lipid peroxidation activity.

3.5. Cell-free redox profiling shows no correlation with protection against oxytosis/ferroptosis in cells

To further understand the mechanism(s) underlying the effects of the compounds on lipid peroxidation, we next assessed the ability of the top 20 anti-oxytotic/ferroptotic compounds to chelate iron and/or directly inhibit lipid peroxidation in cell-free systems in order to exclude effects on other cellular processes.

The term ferroptosis was first introduced to define the iron-dependent generation of lipid peroxides and consequent cell death, and it was shown that iron chelators were able to prevent this toxicity [14, 30]. As such, we began by testing the iron-chelation capacity of the top 20 protective compounds. To do this, the compounds were incubated with ferrous chloride (FeCl₂) and ferrozine solutions. The absorbance of this solution is inversely proportional to the iron (Fe²⁺)-binding capacity of the tested compound. We observed different degrees of Fe²⁺-binding capacity among the compounds but found no correlation with their respective EC₅₀s against oxytosis/ferroptosis (Table 3). In fact, the majority of the top 20 compounds showed little to no ability to chelate Fe²⁺. Hence, the iron-binding capacity of these compounds appears to be unrelated to their ability to prevent oxytosis/ferroptosis in HT22 cells.

Table 3. Cell-free profiling of the top 20 anti-oxytotic/ferroptotic compounds.

LPO (AUC), lipid peroxidation (area under the curve).

Compound	Fe2+-binding capacity (%)	LPO (AUC)
β-lapachone	50.6 ± 7.4	0.97 ± 0.05
DHIT I	2.3 ± 1.1	0.81 ± 0.03
DHT I	6.7 ± 3.4	0.66 ± 0.09
Triptophenolide	6.7 ± 3.0	0.44 ± 0.03
Tanshinone I	3.7 ± 2.2	0.85 ± 0.04
Idebenone	7.6 ± 3.1	0.91 ± 0.05
Cryptotanshinone	71.3 ± 5.3	0.73 ± 0.03
Lipoic acid	7.5 ± 4.5	0.65 ± 0.02
Icaritin	~0	0.55 ± 0.06
Gossypol	8.2 ± 3.7	0.10 ± 0.02
L-Ascorbic acid	3.1 ± 1.5	0.35 ± 0.03
Baicalein	19.3 ± 5.2	0.83 ± 0.02
Catharanthine	40.1 ± 3.0	0.21 ± 0.04
Honokiol	8.7 ± 6.2	0.62 ± 0.02
Carnosol	10.5 ± 3.0	0.14 ± 0.04
3,3’ -Diindolylmethane	8.2 ± 3.7	0.41 ± 0.03
Menaquinone-4	1.1 ± 0.8	0.94 ± 0.15
Dopamine	10.2 ± 4.3	0.27 ± 0.03
Cynaroside	22.4 ± 6.9	0.31 ± 0.03
Fisetin	27.4 ± 6.1	0.17 ± 0.04
Deferiprone	82.7 ± 4.1	-
Trolox	-	-

Given that all of the top 20 protective compounds consistently prevented lipid peroxidation in cells but did not display Fe²⁺-binding capacity, we then asked whether their protection could be a consequence of direct inhibition of lipid peroxidation. To test this, we analyzed the ability of the compounds to inhibit lipid peroxidation in a cell-free assay based on the autoxidation of egg-phosphatidylcholine (PC) liposomes. Unexpectedly, although most of the top 20 protective compounds showed reduced lipid peroxidation in HT22 cells (Fig. 4), their effects on lipid peroxidation in the egg-PC liposomes did not correlate with the cell-based assay, with the best compounds, namely the quinones, presenting the weakest activity (Table 3 and Fig. 5).

Figure 5. Effects on lipid peroxidation in a cell-free liposome-based system.

Co-autoxidation of STY-BODIPY (1.5 μM) and the polyunsaturated lipids of egg-phosphatidylcholine liposomes (1 mM) in the presence of the indicated compounds (10 μM). Lipid autoxidation was initiated using V-70 (0.5 mM) and monitored by measuring the fluorescence increase over time at 37 °C. Graphs show one of n = 3 representative experiments.

The lack of correlation between the effects of the quinones on lipid peroxidation in cells and the cell-free assay prompted us to ask whether the cellular environment could be important for their activity. Quinones are unique chemical entities known to be dynamic redox species in their ability to shift between oxidized and reduced forms and are classified as bioreductive compounds as their activity depends on one- or two-electron reduction catalyzed by various quinone reductases that are present in cells [31–33]. The natural state for most of them is the oxidized form with conjugated cyclic systems [34]. Therefore, the low activity observed for the quinones in the egg-PC-autoxidation cell-free assay could be explained by the absence of the reduced form of the quinones given the lack of oxidoreductases in a cell-free system. To address this possibility, we next tested whether enzymatic reduction of the quinones could alter their anti-lipid peroxidation activity in the cell-free assay.

3.6. Enzymatic reduction of quinones is required for their anti-lipid peroxidation activity in cell-free systems

The ability of NAD(P)H:quinone oxidoreductase 1 (NQO1) to reduce endogenous and exogenous quinones in different models has been known for a long time [35, 36]. More recently, AIFM2, another oxidoreductase that has been renamed as ferroptosis-suppressor protein 1 (FSP1), has been shown to contribute to the anti-oxytotic/ferroptotic activity of coenzyme Q₁₀ (ubiquinone) [37–39], an endogenous quinone.

To determine whether the most protective quinones can be enzymatically reduced in a cell-free environment, the quinones were incubated with recombinant NQO1 or FSP1 using NADH as the immediate electron donor. Since NADH absorbs at 340 nm, the decay of its absorbance was used to determine the enzymatic reduction of the quinones. With the exceptions of tanshinone I and menaquinone 4, all quinones were able to decrease NADH levels over time when incubated with either NQO1 or FSP1 (Fig. 6A).

Figure 6. Enzymatic reduction of quinones increases their anti-lipid peroxidation activity.

(A) The indicated compounds (5 μM) were mixed with NADH (75 μM) alone (CTRL) or with NQO1 or FSP1 (50 nM) and NADH in TBS pH 7.4. NADH levels were monitored by absorbance at 340 nm. (B) Co-autoxidation of STY-BODIPY (1.5 μM) and the polyunsaturated lipids of egg-phosphatidylcholine liposomes (1 mM) in the presence of the indicated compounds (5 μM) and NADH (75 μM) alone (CTRL) or with NQO1 or FSP1 (50 nM) and NADH. Lipid autoxidation was initiated using V-70 (0.5 mM) and monitored by measuring the fluorescence increase over time at 37°C. Graphs show one of n = 3 representative experiments.

To test whether this reduction of the quinones enhances their anti-lipid peroxidation activity in a cell-free system, the quinones were retested in the egg-PC-liposome-based assay in the presence of NQO1 or FSP1 and NADH. All quinones that were substrates of these proteins as identified by the NADH decay assay now displayed an increase in anti-lipid peroxidation activity in the liposomes in the presence of NQO1 or FSP1 (Fig. 6B). However, consistent with the NADH decay assay, the anti-lipid peroxidation activity of tanshinone I and menaquinone 4 was not affected by the presence of the enzymes (Fig. 6B). Since these two quinones both strongly prevent lipid peroxidation in cells, it is possible that they are reduced by other oxidoreductase enzymes [40–42], which may be required for their anti-lipid peroxidation activity.

Altogether, these data support the idea that quinones are a class of chemical compounds with the potential to strongly inhibit oxytosis/ferroptosis by directly targeting lipid peroxidation, but that they require endogenous cellular redox systems for their activity.

Discussion

Our study highlights the importance of plant-derived compounds for the identification and development of novel, potent inhibitors of oxytosis/ferroptosis. Plants display a tremendous array of biochemicals that have evolved to deal with diverse physiological stresses, some of which are relevant to human medicine [17]. Plant primary metabolites (PPMs) are involved in metabolic processes that are required for basic life functions, including cell division, development and reproduction. On the other hand, plant secondary metabolites (PSMs) are not involved in basic functions but rather are produced to deal with abiotic stresses (e.g., UV radiation, drought, heat and soil toxicities) as well as to mediate interactions with antagonists (e.g., herbivores, pathogens and neighboring plants) and mutualists (e.g., fungi, bacteria and pollinators) [43]. Importantly, as in oxytosis/ferroptosis, high levels of lipid peroxides are generated as by-products of plant metabolism, which can occur both enzymatically via LOXs and non-enzymatically by the action of ROS and transition metal ions [17]. In fact, oxytosis/ferroptosis-like cell death has been shown to occur in plants upon stress [44]. These observations are consistent with the high prevalence of PSMs among the best anti-oxytotic/ferroptotic compounds identified in the present study.

To our knowledge, only 9 of the best 20 compounds identified (Cryptotanshinone, idebenone, menaquinone 4, dopamine, lipoic acid, ascorbic acid, baicalein, 3,3’-diindolylmethane and fisetin) have been previously described as anti-oxytotic/ferroptotic molecules [45–56]. Triptophenolide and cynaroside (luteoloside) have been shown to exert anti-oxidant activity [57, 58]. Icaritin, gossypol, carnosol and honokiol have been described as both pro- and anti-oxidant in different studies [59–67]. β-lapachone, DHIT I, DHT I and tanshinone I have been shown to induce ROS production leading to apoptosis and ferroptosis, particularly in cancer cells [68–71].

Four of the best top 20 compounds (DHIT I, DHT I, tanshinone I and cryptotanshinone) are diterpenoid quinones found in the Chinese medicinal herb Danshen (Salvia miltiorrhiza) [72]. Six other compounds in the library (baicalin, danshensu, rosmarinic acid, salvianolic acid, stigmasterol and ursolic acid) are also found in Danshen [72], but offered little to no protection against oxytosis/ferroptosis in HT22 cells. Danshen, a perennial herb native to China, has been traditionally used for centuries to treat cardiovascular diseases [73], but its bioactive constituents possess a variety of pharmacological activities including anti-tumor, anti-oxidant, anti-microbial, anti-viral and anti-inflammatory activities [72]. Importantly, Danshen and several of its constituents have been tested in cellular and animal models of AD and found to be neuroprotective [74, 75]. It will be interesting to investigate in future studies whether there is potential in Danshen to treat AD based on the anti-oxytotic/ferroptotic activity of some of its secondary metabolites, as identified in our screening.

We found that, from all the plant-derived compounds, quinones were enriched and had the highest activity against oxytosis/ferroptosis in the library. Quinones can be found in natural products other than plants, including fungi and bacteria, and have been shown to play a key role in the redox homeostasis of mitochondria and cell membranes [42, 76, 77]. Although recent studies have reported the involvement of ubiquinone in the regulation of oxytosis/ferroptosis [37, 38], the role of other quinones in this process remains poorly understood. Quinones are usually reduced by cellular reductases in either one- or two-electron reductions that require NADH or NADPH [42]. One-electron reduction yields a semiquinone radical (Q^•−), which reacts rapidly with O₂ to form superoxide anion and eventually H₂O₂. Therefore, in the redox environment of biological systems, one-electron reduction of quinones can lead to ROS production [42, 78].

Two-electron reduction yields the corresponding dihydroquinone (QH₂). Several two-electron quinone reductases, including NQO1 and FSP1, have been identified [76, 79, 80]. It is thought that the enzymatic action of NQO1 overrides semiquinone formation by one-electron quinone reductases [79], conferring a quinone-detoxifying role to this enzyme [81]. However, there are also reports showing that NQO1 is necessary to activate quinone toxicity. This property has been investigated as a potential anti-cancer approach using quinones, including β-lapachone and tanshinone, given that NQO1 is frequently overexpressed in a variety of tumors [36, 82–85]. Based on these observations and studies suggesting that the antioxidant role of the QH₂ form of the endogenous ubiquinone is specifically involved in preventing lipid peroxidation versus other types of ROS, it is possible that the QH₂ forms generated by NQO1 and FPS1 are responsible for the anti-oxytotic/ferroptotic activity of the quinones analyzed here via prevention of lipid peroxidation while still leading to an increase in ROS [41, 86–88]. It is likely that, in addition to NQO1 and FPS1, other oxidoreductases participate in the reducing process of the quinones in HT22 cells. This is suggested by the observations with tanshinone I and menaquinone. Future studies should aim at the identification of these enzymes, which evidently will be of value for further understanding the oxytosis/ferroptosis pathway.

In addition, due to the natural tendency of QH₂ for auto-oxidation and the redox cycling that is established between quinone, semiquinone and QH₂, the use of intracellular NADH and NADPH by redox enzymes in this cycle can further exacerbate the production of ROS [89]. The variable instability of the reduced form of some of the quinones analyzed here could explain the lack of correlation between the respective rate of NADH consumption and the observed anti-lipid peroxidation activity in the cell-free systems in the presence of NQO1 or FSP1. Moreover, these observations could explain the increased levels of ROS under basal conditions that we observed upon treatment with the most protective quinones. The fact that these quinones act differently on ROS and lipid peroxidation shows that a compound can be pro-oxidant while still inhibiting lipid peroxidation, highlighting the importance of therapeutically targeting lipid peroxidation in this cell death mechanism. However, although there has been interest in testing quinones as therapeutics against neurodegenerative diseases with positive results, namely idebenone and β-lapachone [90–93], indicating promising therapeutic potential, their use in the clinic for this purpose must be viewed with caution given the possibility of generating ROS, which can damage other molecules (protein, DNA) [94, 95]. It should be added that confirmation of the top 20 compounds as potential AD drug candidates, the ones that have not been studied as such yet, will require proper testing in animal models of the disease.

Several studies suggest that pathological Aβ accumulation in AD can be initiated within neuronal cells [96–99]. Our lab has recently shown that intracellular Aβ accumulation induces oxytosis/ferroptosis in MC65 nerve cells while reducing GSH levels as well as increasing lipid peroxidation and mitochondrial ROS production [13, 14]. In fact, intracellular Aβ accumulation has been associated with the disruption of a variety of cellular functions that overlap with those most affected during oxytosis/ferroptosis such as autophagy or mitochondrial function [100–102]. Our results in MC65 cells further support the idea that the cell death caused by the toxic intracellular accumulation of Aβ can be prevented by treating the cells with compounds that target lipid peroxidation and prevent oxytosis/ferroptosis.

A growing body of evidence shows that features of oxytosis/ferroptosis are also found in other neurological disorders including Parkinson’s disease, Huntington’s disease and amyotrophic lateral sclerosis [3, 4, 103, 104]. For example, decreases in GSH along with iron accumulation and lipid peroxidation have been reported in the regions of the nervous system that are preferentially affected in all of these diseases [103–107]. Notably, iron plays an important role in oxytosis/ferroptosis, as iron chelators prevent oxytosis/ferroptosis-mediated cell death [3, 30]. It is thought that iron can not only generate ROS via the Fenton reaction but also promote the activation of the non-heme iron-containing LOX enzymes [108]. However, our data show that neither iron chelation nor GSH modulation are responsible for the protective effects of most of the anti-oxytotic/ferroptotic compounds identified in this study. In fact, the best compounds are much stronger inhibitors of oxytosis/ferroptosis than the other compounds tested that exhibited iron chelation activity. This observation indicates that phenotypic screening using cells exposed to RSL3 or glutamate is substantially more efficient at identifying potent inhibitors of oxytosis/ferroptosis than selecting compounds based on their capacity to chelate iron.

In addition to the MC65 assay, our lab has been using other cell-based assays for the identification of neuroprotective compounds which represent additional toxicities relevant to dementia [109]. This is the case of energy loss in the brain, which can be mimicked in cultured HT22 cells using IAA [110]. We and others have shown that GSH levels are decreased and lipid peroxidation is enhanced in this cellular model of in vitro ischemia [111, 112]. Our observations that the protection against oxytosis/ferroptosis by the best 20 compounds strongly correlated with the protection against IAA and intracellular Aβ toxicities not only support a role for oxytosis/ferroptosis in these two pathological processes relevant to aging and dementia, but also further reinforce using a phenotypic screening approach based on oxytosis/ferroptosis insults to identify new neuroprotective drugs that have a broad range of beneficial activities.

Conclusions

In summary, our study shows that plants represent an excellent source of natural anti-oxytotic/ferroptotic compounds, and that this potential may favor specific chemical classes that directly target lipid peroxidation in cells. Once identified, these compounds could either be tested as drug candidates, if their physicochemical and safety properties allow, or used as lead compounds for further improvement via medicinal chemistry. Given the growing central role that oxytosis/ferroptosis is being reported to play in a diversity of human diseases that extend beyond the CNS, the value of having new therapeutics that target the pathway cannot be overstated. In addition, our data highlight the necessity for using cell-based assays as opposed to solely relying on cell-free assays, which do not account for the redox systems present in cells that may be required for the activity of certain classes of compounds. Finally, the value of using oxytosis/ferroptosis as a screening strategy is further substantiated by the fact that anti-oxytotic/ferroptotic compounds can also act upon other pathological processes that are critical for aging and neurodegeneration, such as energy loss and Aβ toxicity.

Supplementary Material

1

Table S1. Anti-oxytotic/ferroptotic activity of all the compounds in the library.

2

Table S2. Anti-oxytotic/ferroptotic activity of the best 58 compounds in the library.

Highlights.

• Plants display tremendous potential as source of anti-oxytotic/ferroptotic drugs

• Plant-derived quinones have highest activity among all chemical classes

• Pro-oxidant compounds can prevent lipid peroxidation and oxytosis/ferroptosis

• Anti-lipid peroxidation activity of quinones depends on cellular oxidoreductases

Abbreviations:

Aβ

β-amyloid

ABTS

2,2’-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid)

AD

Alzheimer’s disease

AIFM2

apoptosis inducing factor mitochondria associated 2

ATP

adenosine triphosphate

CNS

central nervous system

Cys

cysteine

DHT I

dihydrotanshinone I

DHIT I

dihydroisotanshinone I

EC₅₀

half maximal effective concentration

FSP1

ferroptosis suppressor protein 1

GAPDH

glyceraldehyde 3-phosphate dehydrogenase

Glu

glutamate

GPX4

glutathione peroxidase 4

GSH

glutathione

HD

Huntington’s disease

IAA

iodoacetic acid

LOX

lipoxygenase

NAD

nicotinamide adenine dinucleotide

NADH

nicotinamide adenine dinucleotide hydrogen

NQO1

NAD(P)H:quinone oxidoreductase 1

PC

phosphatidylcholine

PD

Parkinson’s disease

PK

pharmacokinetics

PPM

plant primary metabolites

PSM

plant secondary metabolites

QH₂

dihydroquinone

Q^•−

semiquinone radical

ROS

reactive oxygen species