Enhanced defense against ferroptosis ameliorates cognitive impairment and reduces neurodegeneration in 5xFAD mice

Abstract

Oxidative damage including lipid peroxidation is widely reported in Alzheimer’s disease (AD) with the peroxidation of phospholipids in membranes being the driver of ferroptosis, an iron-dependent oxidative form of cell death. However, the importance of ferroptosis in AD remains unclear. This study tested whether ferroptosis inhibition ameliorates AD. 5xFAD mice, a widely used AD mouse model with cognitive impairment and robust neurodegeneration, exhibit markers of ferroptosis including increased lipid peroxidation, elevated lyso-phospholipids, and reduced level of Gpx4, the master defender against ferroptosis. To determine if enhanced defense against ferroptosis retards disease development, we generated 5xFAD mice that overexpress Gpx4, i.e., 5xFAD/GPX4 mice. Consistent with enhanced defense against ferroptosis, neurons from 5xFAD/GPX4 mice showed an augmented capacity to reduce lipid reactive oxygen species. In addition, compared with control 5xFAD mice, 5xFAD/GPX4 mice showed significantly improved learning and memory abilities and had reduced neurodegeneration. Moreover, 5xFAD/GPX4 mice exhibited attenuated markers of ferroptosis. Our results indicate that enhanced defense against ferroptosis is effective in ameliorating cognitive impairment and decreasing neurodegeneration of 5xFAD mice. The findings support the notion that ferroptosis is a key contributor to AD pathogenesis.

Graphical Abstract

Introduction

Alzheimer’s disease (AD) is the most common form of dementia and affects millions of people worldwide. There are no effective treatments for AD, as all treatments reduce symptoms but do not slow the underlying progression of the disease [1]. Reactive oxygen species (ROS) are generated during oxidative metabolism, and elevated ROS are observed in AD brains [2, 3]. Polyunsaturated fatty acids (PUFAs) are major components of phospholipids in neurons. PUFA are vulnerable targets of ROS with their oxidation in membrane phospholipids by ROS triggering free radical-mediated damage identified as lipid peroxidation that can lead to cell injury and death. Because increased lipid peroxidation products are readily detected in AD brains [4–7], lipid peroxidation is believed to be a major neural toxicity mechanism [3, 8]. However, AD clinical trials of supplementation of vitamin E, a lipid-soluble antioxidant that inhibits the propagation phase of lipid peroxidation provided limited protection [9–12]. Therefore, the importance of lipid peroxidation in AD pathogenesis remains to be established.

Phospholipid hydroperoxides (PLOOHs) are the primary products of lipid peroxidation. The presence of an excess of PLOOHs in membranes impairs function. PLOOHs are also unstable and, in the presence of redox-active transition medal iron, generate small molecule electrophiles, which elicit sustained free radical-mediated reactions to amplify the initial damage [13]. Notably, a high PLOOHs load triggers ferroptosis, an iron-dependent oxidative mode of cell death that is genetically, morphologically, and biochemically distinct from other modes of cell death such as apoptosis [14, 15].

Glutathione peroxidase 4 (Gpx4) is an antioxidant defense enzyme that reduces PLOOHs in membranes. Reduction of PLOOHs to non-toxic phospholipid alcohols by Gpx4 not only limits damage to the membrane and also blocks the propagation of lipid peroxidation [8]. Because of its highly effective role in reducing membrane PLOOHs, Gpx4 serves as the master regulator of ferroptosis [16]. Notably, the role of Gpx4 in reducing PLOOHs in numerous cell types including neurons is essential, as whole-body or neuron-specific knockout of Gpx4 in mice leads to fetal death [17–19]. Gpx4 is also essential for survival of neuron populations such as forebrain neurons and spinal motor neurons in adult animals, as conditional ablation of Gpx4 induces neuronal death via ferroptosis [20, 21]. The results from studies using mouse models with Gpx4 ablation indicate that alleviation of ferroptotic stress is essential for the health and survival of neurons.

In addition to increased lipid peroxidation, AD brains also display other characteristics of ferroptosis such as iron dysregulation [22]. The importance of ferroptosis to AD pathogenesis, however, remains unknown. 5xFAD mice recapitulate many AD-related features and have a relatively early and aggressive presentation [23]. Notably, unlike other AD mouse models including Tg2576 mice and triple-transgenic (3xTg-AD) mice that show little to no loss of neurons [24, 25], 5xFAD mice exhibit extensive neurodegeneration [23]. Thus, 5xFAD mice appear to be an excellent model for determining the role of ferroptosis in AD. In this study, we tested whether ferroptosis contributes to neurodegeneration in 5xFAD mice and whether ferroptosis inhibition due to enhanced Gpx4 function alleviates the AD phenotypes. To this end, we assessed ferroptosis markers in 5xFAD mice and investigated behavioral function and pathologies of 5xFAD mice with overexpression of Gpx4. The results of this study document that enhanced defense against ferroptosis was effective in improving cognition and suppressing neurodegeneration of 5xFAD mice.

Materials and Methods

Animals and animal procedures

5xFAD mice were purchased from Jackson Labs (C57/B6J background, #34848-JAX). Tg(GPX4) mice (Tg5 line, C57/B6J background) were maintained in house. Male 5xFAD mice were bred with female Tg(GPX4) mice to obtain offspring 5xFAD, 5xFAD/GPX4, WT, and GPX4 mice.

The spatial learning and memory ability of mice was measured by the Morris water maze task, as described previously [21]. In brief, mice were trained to find a hidden platform in opaque water for 5 days with 4 acquisition trials per day from pseudorandomized start positions. Escape latency, the time to find the hidden platform, was recorded as an index of spatial learning and memory. Probe trials were performed on day 6 after acquisition trials, whereby times of platform crossing were recorded and used as an additional memory metric.

Rotarod performance was measured with a Rotamex 4/8 (Columbus Instruments, Columbus, OH) using an accelerating rod protocol [26]. The initial speed of the rod was set to 2 rpm with a linear acceleration to 20 rpm over 300 s. The latency to fall was measured in three trials. The mean latency was registered and used as an indicator of rotarod performance.

A hang wire test was used to evaluate muscle strength [26]. The mouse was placed on the bottom of a wire mesh basket (12”×8”×4”), which was then inverted and suspended above the home cage. The latency to fall was recorded. The mean latency of three trials was used as an indicator of hang wire performance.

Procedures for handling mice in this study were reviewed and approved by the Institutional Animal Care and Use Committees of the University of Texas Health San Antonio and the Audie Murphy Memorial Veterans Hospital, South Texas Veterans Health Care System. The study was carried out in compliance with the ARRIVE guidelines. All methods were performed in accordance with the relevant guidelines and regulations.

Primary hippocampal neuron culture and C11-BODIPY fluorescence microscopy

Primary hippocampal neurons were cultured as previously described [27]. In brief, pregnant day 18 (E18) mice were sacrificed, pups were decapitated, and the forebrains were dissected and transferred in ice-cold Ca²⁺ and Mg²⁺-free Hank’s balanced salt solution (Invitrogen, Carlsbad CA, USA). The hippocampi were dissected out and incubated in 0.05% trypsin-EDTA for 15 min at 37 °C followed by inactivation with 10% fetal bovine serum (FBS) in Neurobasal complete medium. The cells were gently triturated with fire-polished Pasteur pipettes, and then passed through a 70 um cell strainer before seeding. The cells were finally seeded in Poly-D-Lysine (Sigma, St. Louis, MO USA)-coated 6- or 24-well plates in Neurobasal medium (Invitrogen, Carlsbad CA, USA) with 10% FBS, 0.5 mM L-glutamine, and 25 μg/ml penicillin/streptomycin in a humidified atmosphere of 95% air and 5% CO2 at 37 °C. After 8 h of plating, the serum medium was replaced with Neurobasal media with 2% B-27 supplement (Gibco, Waltham, MA USA), 0.5 mM L-glutamine, and 25 μg/ml penicillin/streptomycin. On days in vitro (DIV) 2, the cultures were incubated with 1 uM cytosine arabinoside and maintained for 4 days to limit the growth of glial cells. The media were changed every 72 h until DIV 10.

Lipid ROS levels were determined using BODIPY 581/591 C11 fluorophore in hippocampal neurons. On DIV 12, the neurons were treated with 1 mM t-BuOOH for 1 h and incubated with 5 uM BODIPY C11 fluorophore in culture media for 30 min at 37 °C in a humidified atmosphere of 95% air and 5% CO₂. After incubation, the media was removed, and the neurons were washed with warm Ca²⁺ and Mg²⁺-free Hank’s balanced salt solution (Invitrogen, Carlsbad CA, USA) twice to remove the excessive fluorophore. Subsequently, the neurons were incubated in Hank’s balanced salt solution for imaging. Images were taken with the EVOS M7000 Imaging System (Invitrogen, Carlsbad CA, USA) with an EVOS onstage incubator for precise control of temperature, CO₂, and humidity. Since there is a shift in fluorescence from red to green when C11-BODIPY(581/591) is oxidized, images were taken at 40 x objective using GFP and RFP LED cubes.

Brain tissue section preparation and staining

Mice were anesthetized and then transcardially perfused with PBS followed by 4% paraformaldehyde. Whole perfused brains were collected, post-fixed in 4% paraformaldehyde overnight at 4 °C, and then equilibrated in 30% sucrose in PBS for 3 days at 4 °C. Brains were sectioned at 16 μm using a cryostat (CM1850, Leica, Germany). For Nissl staining, sections were stained with 0.1% (w/v) cresyl violet for 5 min, dehydrated through graded ethanol rinses, and then cleared in xylene.

For immunofluorescence staining, brain sections were re-hydrated in PBS for 10 min and then placed in blocking buffer (10% BSA in PBS and 0.2% Triton X-100) for 1 hr at room temperature. Sections were then incubated in antibody dilution buffer containing primary antibody overnight at 4 °C. Slides were then washed 3 times in PBS followed by 1 hr incubation with secondary antibody conjugated to a fluorophore (ThermoFisher, MA). After washing 3 times, slides were mounted with ProLong Gold mounting media containing 4′,6-diamidino-2-phenylindole (DAPI) (ThermoFisher, MA). Immunofluorescence images were obtained using an Olympus FV-1000 laser scanning confocal microscope (Olympus, PA).

Gpx4 activity measurement

Specific activity of Gpx4 was measured using phosphatidylcholine hydroperoxide (PCOOH) as substrate, as previously described [28]. To synthesize PCOOH, L-α-Phosphatidylcholine (Sigma-Aldrich, St. Louis, MO) was reacted with lipoxygenase (~250,000U) (Sigma-Aldrich, St. Louis, MO) in the presence of oxygen for 30 min. The reaction mixture (i.e., PCOOH mixture) was then applied to a Sep-Pak C18 Cartridge. After washing with water, PCOOH was eluted with Methanol.Methanolic solution of PCOOH was stored at −20 °C and used within 3 weeks.

Cortical tissues were homogenized in lysis buffer (50 mM Tris–HCl [pH 7.5], 0.5 mM EDTA, 1.5% Triton X-100, 0.5 mM DTT, and 1 mM PMSF) supplemented with protease inhibitors. After centrifugation (10,000 × g for 5 min at 4 °C), supernatants were transferred to fresh tubes pre-chilled on ice, and protein concentrations of supernatants were measured. Gpx4 activities in cortical tissue extracts (i.e., supernatants) were measured in Reaction Buffer (0.1 m Tris-HCl [pH8.0), 2 mM EDTA, 1.3 mM NaN3, 0.1% Triton X-100) containing glutathione reductase (1.5 U/ml), glutathione (3 mM), and NADPH (0.2 mM). The reactions were initiated by quickly adding PCOOH (~0.3 mM). Absorbance at 340 nm was measured every 20 seconds over a 10–15 minute period in a spectrophotometer. The Gpx4 activity was expressed as Units/ug protein, with one unit of Gpx4 defined as the oxidation of 1 nmol of NADPH to NADP⁺ per minute at 25 °C.

Glutathione (GSH) assay

Total GSH levels in the cortex tissues were determined using a Glutathione Colorimetric Detection Kit (EIAGSHC, Invitrogen, Carlsbad CA, USA) following the manufacturer’s protocol. The assay was performed using the standard for glutathione provided in the kit. The absorbance was determined at 405 nm by a microplate reader, and the levels of GSH were expressed as nanomoles per gram of tissue protein.

Antibodies and Immunoblotting

The antibodies used in this study are as follows: anti-NeuN, anti-synaptophysin, and anti-β-Actin from Cell Signaling Technology (Beverly, MA); anti-4-HNE antibody from R&D Systems (Minneapolis, MN); anti-GPX4 antibody from Santa Cruz Biotechnology (Dallas, TX); anti-SNAP25, anti-synaptotagmin 1, and anti-BACE1 from Abcam (Cambridge, MA); 6E10 antibody from Covance (Princeton, NJ); Anti-APP antibody from Invitrogen (Carlsbad, CA); and Anti-APP C-terminal antibody from EMD Millipore (Billerica, MA).

Cortical tissues were homogenized in RIPA buffer (20 mM Tris, pH 7.4, 0.25 M NaCl, 1 mM EDTA, 0.5% NP-40, and 50 mM sodium fluoride) supplemented with protease inhibitors (539134, EMD Biosciences Inc., San Diego, CA). Levels of specific proteins in tissues were determined by immunoblotting as previously described [20]. In brief, 30 μg total protein per sample was separated by SDS-PAGE and transferred to PVDF membranes. Membranes were blocked with 5% BSA and then incubated with primary antibody overnight at 4 °C. After incubation with fluorophore-conjugated secondary antibodies (ThermoFisher, MA) for 1 hr, bands were detected using an Odyssey scanner (LI-COR, Lincoln, NE). Alternatively, after incubation with primary antibodies, the membranes were incubated with a HRP-conjugated secondary antibody. The bands were visualized using the ECL Kit (RPN2132, GE Healthcare, Piscataway, NJ). The bands were quantified using NIH ImageJ software (ImageJ 1.52a; http://imagej.nih.gov/ij) and normalized to the loading control (β-Actin).

Real-Time qRT-PCR

Total RNA was isolated with Trizol reagent (Sigma-Aldrich, St. Louis, MO) and then reverse-transcribed using the iScript RT kit (Bio-Rad, Hercules, CA). Real-Time qRT-PCR was conducted as described previously [21]. The mRNA levels were normalized to β-Actin to control for input RNA. Primers used were as follows: Gpx4 (forward: 5’-AGT ACA GGG GTT TCG TGT GC-3’; reverse: 5’-CAT GCA GAT CGA CTA GCT GAG -3’); β-Actin (forward: 5’-ATC TGG CAC CAC ACC TTC TAC-3’; reverse: 5’-CAG GTC CAG ACG CAG GAT G-3’).

Multidimensional mass spectrometry-based shotgun lipidomics (MDMS-SL).

Mouse brains were collected immediately after sacrifice, and cortices from the left brain hemisphere were separately snap-frozen. A small amount of tissue (~10 mg) was homogenized in 0.5 mL of ice-cold diluted PBS (0.1x) with a Potter-Elvehjem tissue grinder. Protein assay for individual homogenates was conducted. An aliquot of homogenate was transferred to a disposable glass test tube. A mixture of internal lipid standards for quantification of all reported lipid classes was added to the tube based on the tissue protein content. Lipid extraction was performed by a modified Bligh and Dyer method as previously described [29]. All of the lipid extracts were flushed with N₂, capped, and stored at −20 °C.

For Electrospray ionization (ESI) direct infusion analysis, lipid extracts were diluted to a final concentration of ~500 fmol/μL, and the mass spectrometric analysis was performed on a Q-Exacutive mass spectrometer (Thermo, San Jose, CA) for cardiolipin, lyso-cardiolipin, and PG analysis and on a QqQ mass spectrometer (Thermo TSQ Altis, San Jose, CA) for analyses of other lipids. Both instruments were equipped with an automated nanospray device (TriVersa NanoMate, Advion Bioscience Ltd., Ithaca, NY) to ionize the lipid species by nano-ESI and operated with Xcalibur software as previously described [30, 31]. Identification and quantification of all lipid molecular species of interest were performed using an in-house automated software program following the principles for quantification by mass spectrometry as previously described [31]. Fatty acyl chains of lipids were identified and quantified by neutral loss scans or precursor ion scans of corresponding acyl chains and calculated using the same in-house software program. All lipid levels were normalized to sample protein content.

Measurement of Aβ by ELISA

Aβ levels were measured as previously described [32]. In brief, the snap-frozen cerebral cortex was homogenized in 2% SDS with protease inhibitor cocktail (EMD Biosciences Inc., San Diego, CA) and sonicated for 1 min. After centrifugation at 200,000 g for 1 h at 20 °C, the supernatant containing SDS-soluble Aβ peptides was collected. The resulting pellet was sonicated in 5 M guanidine and 50 mM Tris-HCl and solubilized by agitation at room temperature for 3–4 h, followed by centrifugation at 13,000 g for 20 min at 4 °C. The supernatant containing SDS-insoluble Aβ peptides was diluted 10-fold to reduce the concentration of guanidine-HCl. Aβ42 and Aβ40 were measured by ELISAs using commercial kits (Invitrogen, Carlsbad, CA). Total protein concentration was measured with a Bradford assay before the ELISAs, and total Aβ levels (SDS soluble plus SDS insoluble) were expressed as pg per mg of total protein.

Statistical analysis

All data are expressed as mean ± SEM. Two-way ANOVA with Tukey’s post-hoc test or Student’s t-test was used to analyze the data. Statistical significance was set to a minimum of p < 0.05.

Results

1. Increased lipid peroxidation and presence of a lipid signature of ferroptosis in 5xFAD mice

5xFAD mouse is a widely used AD mouse model in which amyloid plaques are apparent in the brain of animals as young as two months of age and where progressive neuron loss occurs in multiple brain regions [23, 33, 34]. Ferroptosis is a mode of cell death characterized by its iron dependence. Because dyshomeostasis of iron has been reported in 5xFAD mice [35], we tested whether ferroptosis might play a role in the pathogenesis of 5xFAD mice. Since increased lipid peroxidation is the driver of ferroptosis, we first determined whether 5xFAD mice exhibit increased lipid peroxidation. In 5xFAD mice, cognitive impairments begin at 4 to 5 months of age [23, 36] with severe impairments occurring at around 9 months of age [37]. To determine the status of lipid peroxidation, cortical tissues were collected from pre-symptomatic 5xFAD mice (3 months of age), symptomatic 5xFAD mice (9 months of age), and age-matched WT mice. Immunoblotting was performed to measure tissue levels of 4-hydroxynonenal (4-HNE), a lipid peroxidation end-product and an accepted marker of lipid breakdown [5]. While no differences were observed between pre-symptomatic 5xFAD and WT mice (data not shown), symptomatic 5xFAD mice showed an elevated level of 4-HNE protein adducts compared to age-matched WT mice (Fig. 1A, 1B), indicative of elevated lipid peroxidation.

Fig. 1. Increased lipid peroxidation, elevated lysophospholipids and reduced Gpx4 function in symptomatic 5xFAD mice.

A. Graphs of immunoblots showing 4-hydroxynonenal (4-HNE) protein adducts in cortical tissues from symptomatic 5xFAD mice (9 months of age) and age-matched WT mice. Arrow heads indicate proteins with 4-HNE adducts. B. Quantified results of immunoblots indicating increased 4-HNE level in 5xFAD mice. C. Relative abundances of lipid classes of cortical tissues of 5xFAD and WT mice quantified by multidimensional mass spectrometry-based shotgun lipidomics. PI, Phosphatidylinositol; PS, Phosphatidylserine; CBS, cerebroside; PC, phosphatidylcholine; LPC, lyso-phosphatidylcholine; PE, phosphatidylethanolamine; LPE, lyso-phosphatidylethanolamine; AC, acyl carnitine; CER, ceramide; SM, sphingomyelin; CL, Cardiolipin; PA, Phosphatidic acid; PG, Phosphatidylglycerol; ST, Sulfatide. n = 4, *: p < 0.05. D. Graph showing levels of Gpx4 activity in cortical protein extracts of WT and 5xFAD mice. n = 5 for both groups, *: p < 0.05. E. Graph of immunoblots showing Gpx4 protein in cortical tissues of WT and 5xFAD mice. n = 3 for WT and 5 for 5xFAD. *: p < 0.05. F. Graph showing levels of GSH in cortical tissues of WT and 5xFAD mice. n = 5 for both groups, ns: not significant.

The increased lipid peroxidation in symptomatic 5xFAD mice prompted an investigation into whether changes in lipid composition might have occurred in these mice. We quantified and compared lipid classes and species of WT and 5xFAD mice using multidimensional mass spectrometry-based shotgun lipidomics (MDMS-SL), which profiles lipid classes and analyzes lipid species directly from lipid extracts of biological samples with high accuracy/precision [31, 38]. As described in the Methods, lipids from cortical tissues of symptomatic 5xFAD and age-matched WT mice were extracted, and electrospray ionization (ESI) mass spectrometric analysis was performed to identify and quantify lipid classes and lipid species in the extracts [30, 31]. The relative abundance of each lipid classes in WT and 5xFAD mice are summarized in Fig. 1C. Compared to WT mice, 5xFAD mice showed significantly increased levels of two lysophospholipids: lyso-phosphatidylcholine (LPC) and lyso-phosphatidylethanolamine (LPE). Although PC (phosphatidylcholine), AC (acyl carnitine), and CER (ceramide) were also trending higher in 5xFAD mice than in WT mice, their differences were not statistically significant. LPC and LPE are generated via enzymatic cleavage of one of the two fatty acid tails of phosphatidylcholine and phosphatidylethanolamine, respectively. Increased levels of LPE and LPC are observed in ferroptotic cancer cells after radiation and are believed to serve as a lipid signature of ferroptosis [39]. Thus, in addition to increased lipid peroxidation, symptomatic 5xFAD mice also exhibited a lipid signature of ferroptosis.

Considering the importance of Gpx4 in reducing PLOOHs which suppresses ferroptosis, we also examined whether 5xFAD mice had altered function of Gpx4. To that end, we first compared the activity of Gpx4 between 5xFAD mice and control WT mice. Cortical tissue extracts were obtained from symptomatic 5xFAD mice and age-matched WT mice. The specific activity of Gpx4 was measured using phosphatidylcholine hydroperoxide (PCOOH) as substrate [28]. As shown in Fig. 1D, symptomatic 5xFAD mice showed a decreased Gpx4 activity compared to that in WT mice. We next compared levels of Gpx4 protein using immunoblotting. As shown in Fig. 1E, consistent with the reduced Gpx4 activity, cortical extracts from symptomatic 5xFAD mice had a significantly lower level of Gpx4 protein. Symptomatic 5xFAD mice also had a lower level of total GSH compared to WT mice, but the difference didn’t reach statistical significance (Fig. 1F). To determine whether the deficiency of Gpx4 in 5xFAD mice was due to reduced Gpx4 expression, we further compared Gpx4 mRNA levels between 5xFAD and WT mice by Real-Time qRT-PCR. No significant difference in Gpx4 mRNA was observed between 5xFAD and WT mice (data not shown). Thus, the deficiency in Gpx4 in symptomatic 5xFAD mice appeared to occur at the post-transcription stage.

2. Overexpression of Gpx4 enhanced 5xFAD neurons’ capacity to attenuate lipid ROS

The presence of ferroptotic markers in symptomatic 5xFAD mice persuaded us to determine the importance of ferroptosis in this mouse strain. We thus considered how enhanced defense against ferroptosis might affect the disease phenotypes. Gpx4 plays a key role in defending against ferroptosis by reducing PLOOHs in the membrane; therefore, increased Gpx4 activity would be expected to provide enhanced defense against ferroptosis. We previously generated a Gpx4 transgenic mouse model called Tg(GPX4) mouse. Tg(GPX4) mouse was generated using an endogenous human GPX4 gene, and exhibited overexpression of Gpx4 in a variety of cells and tissues, with a 2–3-fold increase in Gpx4 protein levels in the brain [40]. To generate 5xFAD mice that overexpress Gpx4, we mated Tg(GPX4) mice with 5xFAD mice and obtained double-transgenic offspring harboring both 5xFAD and GPX4 transgenes, designated hereafter as 5xFAD/GPX4 mice.

To ascertain whether 5xFAD/GPX4 mice had increased Gpx4 function, we compared Gpx4 activity in control 5xFAD mice versus 5xFAD/GPX4 mice. As shown in Fig. 2A, 5xFAD/GPX4 mice had a 2.8-fold increase in Gpx4 activity in cerebral cortical tissues over that in control 5xFAD mice. We also compared Gpx4 protein level in cerebral cortical tissues from 5xFAD/GPX4 and control 5xFAD mice by immunoblotting. As shown in Fig. 2B, 5xFAD/GPX4 mice had a significantly higher level of Gpx4 protein.

Fig. 2. Increased capacity to suppress lipid ROS in neurons from 5xFAD/GPX4 mice.

A. Graph showing levels of Gpx4 activity in cortical protein extracts of 5xFAD and 5xFAD/GPX4 mice. n = 5 for both genotypes, *: p < 0.05. B. Graph of immunoblots showing levels of Gpx4 protein in cortical tissues from 5xFAD and 5xFAD/GPX4 mice. C. Representative images of hippocampal neurons from WT, 5xFAD, and 5xFAD/GPX4 mice loaded with C11 BODIPY (581/591) under basal conditions. D. Representative images of hippocampal neurons from WT, 5xFAD, and 5xFAD/GPX4 mice loaded with C11 BODIPY (581/591) after treatment with t-BuOOH (tBH).

To determine how increased expression of Gpx4 might affect the ability of 5xFAD neurons to handle lipid ROS, we cultured primary hippocampal neurons from WT mice, 5xFAD mice, and 5xFAD/GPX4 mice. C11-BODIPY(581/591) is a fluorophore that shifts fluorescence from red to green upon oxidization by lipid ROS and exhibits ratio-metric properties for quantification of lipid ROS in cells [41]. To compare levels of lipid ROS, neurons under basal conditions and stressed conditions were estimated after loading with C11-BODIPY(581/591); stress was induced by treatment of the cells with lipid ROS generator t-BuOOH (tBH) [42]. The red and green fluorescence emitted by C11-BODIPY(581/591was then captured using a fluorescence microscope. As shown in Fig. 2C, under basal conditions 5xFAD neurons exhibited greater green fluorescence compared with WT neurons, and the difference became more apparent after tBH treatment. Thus, 5xFAD neurons had a reduced ability to detoxify lipid ROS. Notably, compared with 5xFAD neurons, 5xFAD/GPX4 neurons showed reduced oxidation of C11-BODIPY(581/591) under both basal and stressed conditions. The findings thus indicated that overexpression of Gpx4 enhanced the capacity of 5xFAD neurons to detoxify lipid ROS.

3. Augmented cognition of 5xFAD/GPX4 mice

5xFAD mice exhibit sex-biased disease phenotype development, with female mice showing a higher level of Aβ production and more advanced pathology than male mice [43, 44]. To determine how increased Gpx4 function might impact disease phenotypes of 5xFAD mice, cohorts of female WT, GPX4, 5xFAD, and 5xFAD/GPX4 mice were generated for cognition assessment. Because no difference in cognitive function was observed between WT mice and GPX4 mice (data not shown, data available upon request), we chose to compare cognitive function of WT, 5xFAD, and 5xFAD/GPX4 mice, with WT mice being used as cognitively normal controls. The spatial learning and memory abilities of cohorts were determined by Morris Water Maze at 5 months of age, an age at which 5xFAD mice reportedly exhibit cognitive dysfunction [45]. As shown in Fig. 3A–B, WT mice found the location of the hidden platform with increased proficiency with training. In contrast, 5xFAD mice showed no reduction in the time interval to locate the hidden platform over 5 days of training, indicating impaired learning. Notably, 5xFAD/GPX4 mice were able to learn as indicated by their decreased time to reach the hidden platform associated with training. On day 5 of acquisition trial training, the difference in time intervals between 5xFAD mice and 5xFAD/GPX4 mice reached statistically significant (p < 0.05, Two-way ANOVA with Tukey’s post-hoc test). In probe trials conducted on day 6, 5xFAD/GPX4 mice had significantly more platform crossings than 5xFAD mice (Fig. 3C). However, 5xFAD mice and 5xFAD/GPX4 mice displayed no significant difference in swim speed (Fig. 3D), indicating that the differences in Morris Water Maze performance were not confounded by locomotor deficits. Taken together, these data suggest that 5xFAD/GPX4 mice had improved spatial learning and memory ability compared with 5xFAD mice.

Fig. 3. Increased learning and memory of 5xFAD/GPX4 mice.

Learning and memory abilities were assessed by Morris Mater Maze (MWM) task. A. Times to locate the hidden platform for 5-month-old 5xFAD, 5xFAD/GPX4, and WT mice over 5 days of acquisition training. B. Representative MWM swim plots on day 5. C. Numbers of platform area crossing in probe trials conducted on day 6. D. Swimming speed in day 5 training. Sex of mice: female. n = 8 for WT, 13 for 5xFAD, and 9 for 5xFAD/GPX4. *: p < 0.05

4. Reduced neurodegeneration in 5xFAD/GPX4 mice

Since the 5xFAD mouse is an AD mouse model which exhibits robust neurodegeneration, we next determined whether overexpression of Gpx4 retarded this degenerative change. 5xFAD mice develop neurodegeneration progressively and show significant neurodegeneration starting at around 9 months of age [34]. We therefore analyzed and compared brain tissues from 9-month-old WT, 5xFAD, and 5xFAD/GPX4 mice. Consistent with prior reports, 5xFAD mice exhibited noticeable loss of neurons in cerebrocortical layer 5 regions compared with WT mice (Fig. 4A). In contrast, there was no obvious reduction in the number of neurons in the same region in 5xFAD/GPX4 mice. We further compared the extents of neuronal loss by neuronal marker protein levels measured by immunoblots. As shown in Fig. 4B–C, neuronal nuclear protein (NeuN), a neuron-specific protein, was lower in in brain of 5xFAD mice compared with WT mice; however, the decrease was significantly attenuated in 5xFAD/GPX4 mice. As a result, 5xFAD/GPX4 mice had a significantly higher level of NeuN than 5xFAD mice. In addition, SNAP-25, a component of the trans-SNARE complex involved in the neurotransmitter release of synaptic vesicles, was diminished in 5xFAD mice compared with WT mice, but again the reduction was also significantly attenuated in 5xFAD/GPX4 mice. Moreover, synaptotagmin 1 (SYT1), a protein involved in synaptic vesicle trafficking and exocytosis, exhibited a similar pattern as observed for SNAP-25 in 5xFAD and 5xFAD/GPX4 mice (Fig. 4B–C). Our findings thus indicate that 5xFAD/GPX4 mice had significantly decreased neurodegeneration compared with 5xFAD mice.

Fig. 4. Reduced neurodegeneration in 5xFAD/GPX4 mice.

A. Graphs of Nissl-stained cortical layer 5 regions of WT, 5xFAD, and 5xFAD/GPX4 mice (9 months of age). B. Immunoblots results showing levels of neuronal marker proteins in cortical tissues of WT, 5xFAD, and 5xFAD/GPX4 mice. SYT1: Synaptotagmin 1. C. Quantified results of immunoblots. n = 4. *: p < 0.05.

5. Reduced lipid peroxidation and attenuated lipid signature of ferroptosis in 5xFAD/GPX4 mice

To determine if the ameliorated neurodegeneration of 5xFAD/GPX4 mice was due to attenuated ferroptosis, we next compared levels of lipid peroxidation between 5xFAD and 5xFAD/GPX4 mice. Cortical protein extracts from WT, 5xFAD, and 5xFAD/GPX4 mice were probed with anti-4-HNE antibody by immunoblotting (Fig. 5A). Quantified levels of 4-HNE adducts are presented in Fig. 5B. Compared to 5xFAD mice, 5xFAD/GPX4 mice had a significantly reduced level of 4-HNE, indicating that overexpression of Gpx4 resulted in attenuation of lipid peroxidation. We further profiled and compared lipid classes among 9-month-old WT, GPX4, 5xFAD, and 5xFAD/GPX4 mice by MDMS-SL. Among all the lipids, lysophospholipids LPC and LPE showed the most marked changes. As shown in Fig. 5C, overexpression of Gpx4 had no effect on LPC, as no difference was observed between 5xFAD and 5xFAD/GPX4 mice or between WT and GPX4 mice, even though LPC was elevated in both 5xFAD and 5xFAD/GPX4 mice compared with non-5xFAD mice (i.e., WT and GPX4 mice). However, 5xFAD/GPX4 mice had a significantly lower level of LPE than 5xFAD mice, and the LPE level was also depressed in brain of GPX4 mice than in WT mice. The lower LPE levels in both 5xFAD/GPX4 and GPX4 mice thus indicated that overexpression of Gpx4 reduced LPE level specifically. The reduced LPE level of 5xFAD/GPX4 mice thus represented an attenuation of ferroptotic lipid signature afforded by Gpx4 overexpression.

Fig. 5. Ameliorated ferroptotic markers in 5xFAD/GPX4 mice.

A. Graph of immunoblots showing levels of 4-HNE protein adducts in cortical tissues of WT, 5xFAD, and 5xFAD/GPX4 mice. B. Quantified results of immunoblots. *: p < 0.05. C. D. Levels of LPC and LPE in cortical tissues of WT, GPX4, 5xFAD, and 5xFAD/GPX4 mice quantified by MDMS-SL. n = 4. *: p < 0.05.

6. Reduced Aβ levels in 5xFAD/GPX4 mice

In 5xFAD mice, the expression of mutated APP and PS1 increases production of Aβ, particularly Aβ42, leading to robust amyloid plaque formation [23]. To test whether overexpression of Gpx4 affected amyloidosis, brain sections from 9-month-old 5xFAD and 5xFAD/GPX4 mice were stained for amyloid plaques by immunofluorescence. Compared to 5xFAD mice, 5xFAD/GPX4 mice showed reduced presence of amyloid plaques (Fig. 6A). To further compare levels of Aβ burdens, total proteins were extracted from frontal cortex tissues, and levels of Aβ42 and A40, two major species of Aβ, were measured using commercial kits. As shown in Fig. 6B, 5xFAD/GPX4 mice had significantly lower levels of both Aβ42 and Aβ40 than 5xFAD mice. However, no difference in the ratio of Aβ42 to Aβ40 was observed between 5xFAD and 5xFAD/GPX4 mice (data not shown).

Fig. 6. Reduced Aβ levels in 5xFAD/GPX4 mice.

A. Representative brains sections stained with anti-Aβ antibody (6E10) showing amyloid plaques in cortical and hippocampal regions of 5xFAD and 5xFAD/GPX4 mice. B. Aβ levels in frontal cortex tissues of 5xFAD and 5xFAD/GPX4 mice determined by ELISA. n = 9 for 5xFAD and 7 for 5xFAD/GPX4. C. Graphs of immunoblots showing BACE1 and C20 levels in the cortex of WT, 5xFAD, and 5xFAD/GPX4 mice. D. Quantified results of BACE1 and C20 levels. n = 3 for both genotypes. *: p < 0.05

BACE1 is responsible for β-site cleavage of APP during Aβ production. Because prior reports indicated that BACE1 expression is influenced by lipid peroxidation [32, 46], we compared levels of BACE1 protein between 5xFAD and 5xFAD/Gpx4 mice. A statistically significant decrease of BACE1 protein was observed in 5xFAD/GPX4 mice compared with 5xFAD mice (Fig. 6C–D). Consistent with reduced BACE1 protein, the level of APP CTF99, carboxyl terminal fragments (CTF) of APP generated by β-site cleavage [47], was also reduced in 5xFAD/GPX4 mice (Fig. 6C–D). The reduced BACE1 protein and decreased APP CTF99 suggest that the reduced BACE1 expression contributed to the depressed Aβ levels in 5xFAD/GPX4 mice.

7. Increased locomotor function of aged 5xFAD/GPX4 mice

In addition to cognitive impairment, both male and female 5xFAD mice develop motor impairments starting at about 9 months of age, with marked motor impairments being observed in these mice when they achieved an advanced ages (12–16 months) [48, 49]. To determine whether overexpression of Gpx4 would influence motor impairments during aging, we compared locomotor function of cohorts of WT, 5xFAD, and 5xFAD/GPX4 mice at 15 months of age using both a rotarod task and a hang-wire task. Consistent with previous reports, 15-month-old 5xFAD mice had a dramatic decline in both rotarod and hang-wire performance (Fig. 7A–B). Compared to 5xFAD mice, aged 5xFAD/GPX4 mice showed significantly improved performance in both rotarod and hang-wire tasks (Fig. 7A–B). In addition to the reduced locomotor function, aged 5xFAD mice also exhibited a significant loss of bodyweight. Notably, 5xFAD/GPX4 mice showed a significantly higher bodyweight compared to 5xFAD mice (Fig. 7C). Thus, overexpression of Gpx4 not only increased locomotor function but also ameliorated bodyweight loss of 5xFAD mice at advanced age.

Fig. 7. Increased motor function and reduced weight loss of aged 5xFAD/GPX4 mice.

A. Performances of 15-month-old WT, 5xFAD, and 5xFAD/GPX4 mice in the hang-wire task. B. Performances of 15-month-old WT, 5xFAD, and 5xFAD/GPX4 mice in the rotarod task. C. Bodyweight of WT, 5xFAD, and 5xFAD/GPX4 mice at 15 months of age. Sex of mice: male. n = 7 for WT and 6 for both 5xFAD and 5xFAD/GPX4. *: p < 0.05.

Discussion and Conclusions

Increased levels of lipid peroxidation have been well-documented in AD patients [4–7]. However, the role of ferroptosis in the pathogenesis of AD is unclear. AD mouse models with elevated Aβ deposition such as Tg2576 transgenic and triple-transgenic (3×Tg-AD) mice exhibit increased lipid peroxidation. Since no significant neurodegeneration is observed in Tg2576 transgenic and triple-transgenic (3×Tg-AD) mice [24, 25], it is unlikely that ferroptosis plays a major role in these AD models. In this study, we showed that symptomatic 5xFAD mice had significantly elevated lipid peroxidation products, as evidenced by the increased 4-HNE level. Because lipid peroxidation is the driver of ferroptosis, this result implicates the involvement of ferroptosis in neurodegeneration of 5xFAD mice. At present, the characterization of ferroptosis in vivo is hampered by a dearth of specific markers for this destructive process. In our study, we identified major lipid classes and lipid species in symptomatic 5xFAD mice using a lipidomics approach, and observed that 5xFAD mice had significantly increased levels of two lysophospholipids, LPC and LPE. Increased levels of lysophospholipids such as LPC and LPE are associated with ferroptosis [50] and are believed to be a lipid signature of ferroptosis [39].

The current findings demonstrate that, in addition to elevated lipid peroxidation, symptomatic 5xFAD mice also exhibit a lipid signature of ferroptosis consisting of increased LPC and LPE levels. Although apoptotic features such as activation of caspase-3 were previously reported in 5xFAD mice [34], the findings from this study indicate that ferroptosis also contributes to disease pathogenesis of 5xFAD mice. We also report that symptomatic 5xFAD mice had a deficiency of Gpx4, which would result in a compromised anti-ferroptosis defense. Consistently, reduced Gpx4 protein levels were also observed in another APP mouse model [51]. These findings suggest that downregulation of Gpx4 may be a common consequence of Aβ production and accumulation. Moreover, Bourassa et al. (2013) reported that 5xFAD mice had elevated brain iron levels, which are directly linked to ferroptosis [35]. Because of the accumulated lipid peroxidation products, excessive iron concentrations, and robust neurodegeneration, 5xFAD mice appear to be a useful model for determining the role of ferroptosis in AD. It is unclear, however, why ferroptosis may be particularly important in 5xFAD mice. Although 5xFAD mouse model is suitable for modeling cognitive and behavioral deficits at early-, middle-, and late-stage disease progression of AD [52], the neurodegenerative features in the 5xFAD mouse model develop significantly faster than those of other AD models. It is possible that 5xFAD mice may have elevated ferroptotic stress such as a higher load of PLOOHs and/or a greater extent of iron-dyshomeostasis than other mouse models. Follow-up research is required to determine whether this is the case.

Lipid peroxidation has long been hypothesized to be a potential target for AD intervention [3, 5, 53, 54]. However, randomized placebo-controlled trials of using vitamin E supplementation as a treatment for AD have yielded contradictory and inconsistent results: a positive outcome of delayed clinical progression was reported in two trials [9, 10], but no and a worsening effect was reported in two others [11, 12]. The results of vitamin E trials may have been confounded by the limited availability of vitamin E to cross the blood-brain barrier [55], as well as the potential side effects of vitamin E [56]. A reduction of neural lipid peroxidation has also been found to improve cognition and reduce Aβ loads in AD animal models [32, 57, 58]; however, most of the studies were performed in AD mouse models with little or no loss of neurons. The current study used a genetic approach of Gpx4 overexpression to suppress lipid peroxidation in a rather aggressive AD mouse model which exhibits robust neurodegeneration.The results confirm the efficacy of increased defense against lipid peroxidation in ameliorating cognitive decline and reducing Aβ. More importantly, the findings document that a reduction in neural lipid peroxidation is an effective treatment in retarding neurodegeneration, thereby highlighting the potential of ferroptosis as a new target of intervention for AD.

We also demonstrated that 5xFAD/GPX4 mice had a decreased Aβ load, and that reduced BACE1 function may be responsible for the lowered Aβ burden. Our findings are consistent with prior studies showing that Aβ production and/or Aβ accumulation are influenced by the degree of lipid peroxidation [32, 46, 58, 59]. The decreased Aβ load in 5xFAD/GPX4 mice prompts the question as to whether the ameliorated neurodegeneration was due to the reduced accumulation of Aβ. However, Aβ accumulation per se is not overtly toxic in rodents, as many AD mouse models with relatively high Aβ loads do not show significant neurodegeneration [24, 25]. Because 5xFAD/GPX4 mice had only a minor reduction of Aβ burden, we believe that the limited neurodegeneration observed in 5xFAD/GPX4 mice is more likely a result of the suppression of ferroptosis, rather than being related a reduced Aβ level.

In summary, this study showed that symptomatic 5xFAD mice exhibited signatures of ferroptosis, and that increased defense against ferroptosis conferred by Gpx4 overexpression was effective in improving behavior function and reducing neurodegeneration of 5xFAD mice. Our results indicate that augmenting the defense system against ferroptosis may be beneficial for reducing the progression and severity of AD.

Highlights:

1. Symptomatic 5xFAD mice exhibit markers of ferroptosis.

2. 5xFAD neurons’ capacity to detoxify lipid ROS is enhanced by Gpx4, the master defender against ferroptosis.

3. 5xFAD mice with overexpression of Gpx4 show improved cognition and ameliorated neurodegeneration.

4. Markers of ferroptosis are attenuated in 5xFAD mice with overexpression of Gpx4.

Data Availability Statement

Data available on request from the authors.