Aberrant Mitochondrial Metabolism in Alzheimer's Disease Links Energy Stress with Ferroptosis

Abstract

Alzheimer's disease (AD) is defined by β-amyloid plaques and tau-containing neurofibrillary tangles, but the ensuing cellular derangements that culminate in neurodegeneration remain elusive. Here, a mechanistic link between two AD pathophysiological hallmarks: energy insufficiency and oxidative stress is revealed. It is demonstrated that mitochondrial function and glutathione (GSH) flux are coupled, impacting neuronal ferroptosis susceptibility. Analysis of proteomic data from the inferior temporal cortex of 625 subjects along a continuum of clinical and pathological changes in AD, reveals a prominent depletion of mitochondrial proteins. Biogenetic insufficiency in AD is reflected by a concurrent loss of GSH, which requires 2 ATP for its synthesis, and genetic and pharmacologic ATP depletion models confirm that ATP is rate-limiting for GSH. Accordingly, an unbiased association analysis uncovers mitochondrial proteins in positive correlation with total GSH (t-GSH) in AD subjects. But mitochondria also consume GSH via the SLC25A39 transporter. It is found that mitochondrial inhibition either increases or decreases ferroptosis susceptibility in cellular models, depending on contextual factors that dictate whether mitochondria act as a net GSH producer or consumer, respectively. Mitochondria therefore control GSH flux, and loss of energy output is consequently demonstrated as a liability for ferroptosis in AD.

Keywords: ATP; alzheimer's disease; bioenergetics; ferroptosis; glutathione; mitochondria; neurodegeneration.

Figure 1

Large proteomic interrogation of inferior temporal gyrus tissue from 625 MAP study participants. a) Cohort metadata overview. The heatmap shows measures of AD pathology, APOE ε4 genotype, cognitive function, sex (rows) for the 625 study participants (columns), ordered by plaque, study cohort composition and proteomics work‐flow including a WGCNA, which identified the yellow module (MEyellow). Pie charts summarise subject grouping by dementia and pathology (based on the CERAD criteria), ApoE ε4 status and sex. b) cytoscape diagram of proteins in the yellow module with highlighted hub genes and c) MonaGO chord plot of cellular component identification of mitochondria as the yellow module; 1. mitochondrion; 2, mitochondrial inner membrane; 3, mitochondrial matrix; 4, respiratory chain; 5, respiratory chain complex I; 6, mitochondrial intermembrane space; 7, membrane raft, intercalated disk; 8, pyruvate dehydrogenase complex; 9, mitochondrial alpha‐ketoglutarate dehydrogenase complex; 10, mitochondrial crista; and 11; postsynaptic density. d) boxplot showing MEyellow eigenvalue score across MAP study participants grouped by dementia and pathology (based on the CERAD criteria) ND‐ve; n = 165), people who died without dementia but met criteria for AD pathology (ND +ve; n = 213) and people who were diagnosed with AD dementia, and had AD pathology confirmed post mortem (AD+ve n = 237); each dot represents an individual subject; one‐way ANOVA. e) global cognition median split high versus low MEyellow over 5 years preceding death (all subjects; n = 615). Schematics in a created with BioRender.com.

Figure 2

ATP is linked to GSH synthesis revealing ferroptosis vulnerability under energy stress. a) Schematic displaying that two ATP molecules generated by the electron transport chain (ETC) are required to synthesise 1 GSH molecule; γ‐GCS, γ‐glutamylcysteine synthetase; GS, GSH synthetase. b) Time course of t‐GSH depletion with buthionine sulphoximine (BSO) in HT22 neuronal cells. c) boxplot showing t‐GSH across MAP study subjects assigned according to dementia and pathology (CERAD); each dot represents individual subject; a one way ACNOVA with age at death, APOE ε4, and sex as covariates. d) global cognition median split high versus low t‐GSH over 5 years preceding death (all subjects; n = 615). e–i) GSH recovery after 4 h of BSO treatment – GSH, ATP and viability were measured in HT22 cells after 17 h of recovery in the presence of escalating doses of ETC inhibitors (rotenone – complex 1 inhibitor; TTFA – complex II inhibitor; antimycin A – Complex III inhibitor; sodium azide – Complex IV inhibitor; CCCP – mitochondrial uncoupler). j–l) Bar plots of ATP, t‐GSH and viability after 17 h co‐treatment with 5% of either ATP‐encapsulated or empty liposomes with ETC inhibitors; antimycin (250 nm), azide (7.5 mm) and CCCP (3.125 µm). m) schematic diagram showing the action of ATP nucleosidase (Cap 17) cleaving the N‐glycosidic bond between the adenine and sugar moieties of ATP, resulting in adenosine and ribulose 5‐triphosphate products. n) bar plots of ATP and t‐GSH in HT22 cells FACS‐sorted for high transfection of Cap17 (pCap17) or empty vector (pEV). Data in j‐l & n presented as mean ± SEM of 3 independent experiments, multiple two‐sided t‐tests, ^*indicates significance p<0.05. nd indicates no difference. Schematics in a & m created with BioRender.com.

Figure 3

Unbiased analysis of Glutathione in AD reveals a mitochondrial relationship and ferroptosis vulnerability with mitochondrial reduction. a) Workflow schematic of differential protein expression in the MAP proteomics comparing high versus low total GSH, revealing 0 overlap of hits and multiple linear model analysis revealing 162 differentially correlated proteins of which 29 are mitochondrial. b–d) 29 stacked linear models showing the relationship of 29 mitochondrial proteins (Gene Z scores) with total GSH in AD+ve (red, n = 237), ND+ve (yellow, n = 213), and ND‐ve, pathology‐ve (blue, n = 165). e) dot plot of protein expression (Gene Z score) of 29 mitochondrial proteins across MAP subjects assigned according to dementia and pathology (CERAD). f) Viability of HT22 neuronal cells (assayed by Calcein AM) co‐treated with erastin for 17 h in the absence or presence of electron transport chain inhibitors (antimycin A – Complex III inhibitor (green, 200 nm); sodium azide – Complex IV inhibitor (orange, 7.5 mm); CCCP – mitochondrial uncoupler (purple, 3.125 µm)). g) total GSH depletion with erastin (black) in the presence of ETC inhibitors. h) Bar plot of SLC25A39 protein expression (relative to control) in HT22 cells treated in the presence or absence of erastin (10um) with ETC inhibitors, one way ANOVA, ^**** p < 0.0001. i) Bar plot of SLC25A39 protein expression (relative to control) in HT22 cells treated in the presence or absence of RSL3 (1um) + liproxstatin (LPX, 1um) one way ANOVA, ns = not significant. j) Representative western blot images. k–m) viability curves of RSL3 with ETC inhibitors (antimycin A – Complex III inhibitor (green, 200 nm); sodium azide – Complex IV inhibitor (orange, 7.5 mm); CCCP – mitochondrial uncoupler (purple, 3.125 µm) with or without liproxstatin (LPX, 1um) and n–p) with or without glutathione ethyl ester (GSHEE, 1 mm). q, schematic showing the bimodal relationship between mitochondrial protein expression and glutathione created using BioRender.com. k,p data are means± SEM.

Figure 4

Low mitochondria and high iron models an increased ferroptosis risk present in the AD population. a–c) viability curves cells treated with arachidonic acid and increasing doses of iron in the absence and presence of ETC inhibitors (antimycin A – Complex III inhibitor (green, dose); sodium azide – Complex IV inhibitor (orange, dose); CCCP – mitochondrial uncoupler (purple, dose), and with or without liproxstatin (lpx, 1um) d) Quadrant scatter plot using median values of Mitochondria (MEyellow) and iron from MAP subjects: high mitochondria, low iron (Q1), high mitochondria, high iron (Q2), low mitochondria, low iron (Q3) and low mitochondria, high iron (Q4). Each dot represents a participant (AD+ve = red, ND+ve = yellow, ND‐ve = blue). e) Bar plot of percentage of subjects per quadrant according to diagnosis and pathology status. f) Bar plot of total GSH of subjects per quadrant according to diagnosis and pathology status; p values generated using planned comparisons from a one way ANOVA g) global cognition (95%CI) according to quadrant assignment over 5 years preceding death. h) schematic depicting bimodal relationship between mitochondria as GSH consumer and producer i) t‐SNE Gene map from brainscope.nl; an online interface compiling the co‐expression of transcripts across numerous subjects and brain regions using the Alan Brain atlas highlighting glutathione peroxidase 4 (GPX4) in blue with the top 50 co‐expressed genes. j) Bar plot of gene count from pathway enrichment of the top 50 genes clustered with GPX4. k) Bar plot of oxidative phosphorylation gene count from pathway enrichment of antioxidant enzymes (Catalase, CAT; glutathione peroxidase 1, GPX1; glutathione peroxidase 2, GPX2; glutathione peroxidase 3, GPX3; glutathione peroxidase 4, GPX4; NAD(P)H quinone dehydrogenase 1, NQO1; Superoxide dismutase 1, SOD1; Superoxide dismutase 2, SOD2; Superoxide dismutase 3, SOD3; thioredoxin reductase 1, TNXRD1.