Main

Mitochondria couple metabolism and respiration to energy production, but also produce reactive oxygen species (ROS) via partial reduction of oxygen. Mitochondria are key contributors to total ROS levels and their impairment or cell stress can dramatically increase ROS production1. Mitochondria generate ROS (mtROS) from at least 11 distinct sites, with respiratory complex I (CI) and complex III (CIII) considered the major sources and potentially linked to distinct signalling mechanisms and disease states1,2, but exact contributions by specific sites remain unclear. CIII has a high capacity for ROS production due in part to its ubiquitous and high expression2. CIII is also topologically poised to influence intracellular signalling by generating ROS away from the inner-mitochondrial matrix, unlike CI and most other sites1.

The exact regulators and effects of site-specific mtROS production in health and disease remain undefined due to long-standing challenges, including off-target effects of most genetic and pharmacologic manipulations of mtROS and the direct dependence of mtROS on metabolism. To address these challenges, we previously identified potent, small-molecule suppressors of electron leak (SELs), including suppressors of complex IIIQo electron leak (S3QELs or ‘sequels’) and suppressors of complex IQ electron leak (S1QELs or ‘cycles’), which selectively block ROS production from a single mtROS site without altering normal mitochondrial functions like ATP production3,4. SELs are efficacious in diverse systems, including cultured cells, invertebrates and mice3,4,5,6, and uniquely enable the dissociation of site-specific mtROS production from other metabolic processes.

In the brain, mitochondrial dysfunction is an early and prevalent feature of neurodegenerative disorders, including Alzheimer’s disease (AD), frontotemporal dementia (FTD), amyotrophic lateral sclerosis and Parkinson’s disease7, and mtROS are implicated as central, feed-forward drivers of cell dysfunction and neuropathology8,9. Notably, a dominantly inherited mutation in the CIII subunit UQCRC1 is associated with parkinsonism and polyneuropathy and linked to increased basal ROS production in human neural cells10. Similarly, mice with brain-specific deficiency in RISP, the CIII subunit necessary for carrying electrons away from the ROS production site, have increased oxidative stress and early mortality, suggesting that increased CIII ROS levels are sufficient to cause neuropathology11. Indeed, the main contributing factors to many neurodegenerative disorders, including ageing, amyloid-β (Aβ) accumulation, tauopathy and neuroinflammation, are linked to mitochondrial dysfunction and increased mtROS, suggesting a central role for mitochondrial oxidative mechanisms in disease12. In support, cognitive resilience to dementia-related pathology is associated with lower expression of proteins linked to oxidative mechanisms13. Yet, despite mounting evidence implicating mtROS in disease, the exact triggers, temporal dynamics, sources and downstream oxidation targets of mtROS and their contributions to cell signalling and disease mechanisms remain unclear.

Here, we used SELs along with multiple other pharmacological and genetic manipulations to establish that astrocytic CIII ROS are induced in a temporally defined manner by stimuli linked to disease and promote transcriptional programs and neuroimmune cascades that amplify disease pathogenesis. We conclude that CIII ROS represent key signal transducers in astrocytes that can be selectively suppressed to alleviate disease-associated changes in mitochondrial redox signalling and immunometabolic pathways.

Results

Select stimuli trigger astrocytic complex III-derived ROS through NF-κB and NCLX activity

Astrocytes are prevalent glial cells with essential roles in central nervous system (CNS) function and are increasingly implicated in dementia-linked pathogenesis and neuroimmune cascades14,15. Astrocytes have a high capacity for mtROS production, including from CI and CIII (refs. 4,16). Manipulation of astrocytic redox balance and intracellular ROS influences brain metabolism, neuronal function and cognition14,17,18,19, and dysregulation of astrocytic redox pathways may promote neuropathology20,21,22. Cytokines and disease-linked factors can increase astrocytic cellular ROS but exact sources of ROS are uncertain23,24,25.

To investigate the induction and regulation of astrocytic ROS in response to pathogenic stimuli, we first measured the rates of cellular H2O2 efflux in primary mouse astrocytes upon stimulation with neuroimmune factors induced in neurological disease23,26,27,28. The cultures were confirmed to be highly enriched for astrocytes (> 95%; Extended Data Fig. 1a,b). The cytokines interleukin (IL)-6 and interferon (IFN)γ had no effects on H2O2 efflux, whereas a cocktail of cytokines previously linked to astrocyte-dependent neurotoxicity, consisting of IL-1α, tumour necrosis factor (TNF) and complement component 1q (C1q)26, increased the rates of astrocytic H2O2 production by 24 h (Fig. 1a). The cocktail-induced H2O2 production was attributable to IL-1α, but not to TNF or C1q (Fig. 1a), suggesting that astrocytic ROS is engaged by specific types of ligands and signalling mechanisms.

Fig. 1: Astrocytic CIII ROS are induced by specific stimuli.

a–c, Cellular H2O2 efflux after 24 h treatment with vehicle, IL-1α (3 ng ml−1), TNF (30 ng ml−1), C1q (400 ng ml−1), IL-6 (33 ng ml−1), IFNγ (10 ng ml−1) with or without S3QEL2 (20 μM) or S3QEL1.2 (1 µM). n = 6–12 wells, one-way analysis of variance (ANOVA) with Dunnett’s (a) or Tukey’s test (c). d, Schematic of mtHyPer7 sensing mitochondrial CIII-derived H2O2 created with Biorender.com. e, Measurement of mtHyPer7 fluorescence 6 h after treatment with vehicle, IL-1α, S3QEL1.2 (3 μM) or S1QEL2.2 (10 μM). n = 162–310 cells from 5–19 wells, Kruskal–Wallis with Dunn’s test. f, Time-course of mtHyPer7 fluorescence after vehicle or IL-1α. n = 56–296 cells from 3–9 wells, Mann–Whitney, unpaired two-tailed U-test. g, Representative images of mtHyPer7 oxidized/reduced ratios after 6 h treatment with vehicle, IL-1α or IL-1α with S3QEL1.2. Scale bars, 10 µm; 5 µm (inset). h, Time-course of mtHyPer7 fluorescence after vehicle or oAβ (300 nM). n = 54–310 cells from 2–12 wells, Mann–Whitney, unpaired two-tailed U-test. i, Quantification of mtHyPer7 after 3 h treatment with vehicle, oAβ or S3QEL1.2 (3 μM). n = 110–233 cells from 5–6 wells, Kruskal–Wallis with Dunn’s test. j, Representative images of mtHyPer7 oxidized/reduced ratios after 3 h treatment with vehicle, oAβ or oAβ with S3QEL1.2. Scale bars, 10 µm, 5 µm (inset). Data are shown as mean ± s.e.m. Some error bars in f and h are not visible because the error range is smaller than the symbol size. *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant. Additional statistical details are included in Supplementary Table 6.

Source data

To determine the contribution of CIII ROS in astrocyte H2O2 efflux, we treated astrocytes with IL-1α and one of two structurally distinct S3QELs (Fig. 1b) to rule out potential off-target effects. S3QEL1.2 and S3QEL2 inhibited IL-1α-induced H2O2 production to a similar extent (54.7% and 56.3%) (Fig. 1c), confirming the involvement of CIII ROS. In addition, two structurally distinct S1QELs revealed a contribution of CI-derived ROS in these responses (Extended Data Fig. 1c,d), indicating that both CI and CIII contribute to astrocytic ROS production. We also tested inhibitors of several other ROS producing enzymes to account for the remaining S3QEL- and S1QEL-insensitive H2O2 efflux. Inhibition of NOXs with APX-115, DHODH with BAY-2402234, or MAO-B with selegiline did not influence astrocytic H2O2 efflux (Extended Data Fig. 1e–g). Other potential mitochondrial and non-mitochondrial sources remain to be explored (for example, OGDH, SDH, mGPDH, peroxisomal and ER enzymes). To rule out off-target metabolic effects of SELs, ATP levels were measured in astrocytes metabolically restricted to pyruvate- and galactose-dependent respiration. As reported in other cell types3,4,6,29, S3QELs and S1QELs did not affect mitochondrial metabolism (Extended Data Fig. 1h,i).

To further define the mechanisms regulating CIII ROS, we targeted the highly sensitive and selective ratiometric H2O2 sensor HyPer730 to mitochondria (mtHyPer7; Fig. 1d and Extended Data Fig. 2a,b) and performed live-cell confocal imaging. We confirmed that mtHyPer7 is oxidized in response to treatment with antimycin A, a potent inducer of CIII ROS production. Antimycin A-induced oxidation was dose-dependently suppressed by S3QEL1.2 (Extended Data Fig. 2c,d), indicating that mtHyPer7 enables precise and spatially resolved measurements of astrocytic CIII ROS.

Live-cell imaging of mtHyPer7 revealed that S3QEL1.2 had no effects on baseline H2O2 levels, whereas the CI ROS suppressor S1QEL2.2 decreased baseline mitochondrial H2O2 levels (Fig. 1e), suggesting that CI ROS play a greater role than CIII ROS in setting the basal redox tone in astrocyte mitochondria, in line with previous reports4,16. To better understand the temporal dynamics of mtROS induction, we performed confocal imaging of mtHyPer7 at different time points after stimulation with IL-1α. We found that mitochondrial H2O2 was increased within 4 h, peaked by 6 h, and remained elevated until at least 8 h (Fig. 1f). In contrast to its lack of effect on basal mtHyPer7 oxidation, S3QEL1.2 inhibited IL-1α-induced responses by approximately 49% (Fig. 1e,g). Conversely, suppression of CI ROS with S1QEL2.2 did not markedly affect IL-1α-induced mtROS at 6 h (Fig. 1e). Indeed, relative mtROS levels induced by IL-1α with or without S1QEL2.2 were nearly identical (79.8% for IL-1α versus vehicle; 78.5% for IL-1α with S1QEL2.2 versus S1QEL2.2 alone). As a control, we confirmed that the same dose of S1QEL2.2 was fully effective in blocking mtROS induced by the CI-ROS-producing redox cycler MitoParaquat6 (Extended Data Fig. 2e). Notably, co-treatment with S3QEL1.2 and S1QEL2.2 effectively blocked the majority of detectable mtROS (Extended Data Fig. 2f), revealing the underlying additive contributions of CI ROS and CIII ROS. Inhibition of NOXs with APX-115 did not affect IL-1α-induced mtROS after 6 h (Extended Data Fig. 2g) and IL-1α did not increase astrocytic NOX levels at this time (Extended Data Fig. 2h). Together, these results suggest that CI ROS production contributes to astrocytic mtROS levels in a sustained manner, whereas CIII ROS production is selectively recruited upon cell stimulation with specific factors.

In addition to IL-1α, we examined the effects of oligomeric Aβ (oAβ), which is strongly implicated in AD and may cause pathology in part through its effects on astrocytes and increasing ROS levels7,31. Indeed, we found that acute treatment with oAβ caused dynamic changes, increasing astrocytic mtROS within 2 h, which remained elevated until 3 h after stimulation (Fig. 1h). The onset and recovery of oAβ effects were faster and more dynamic compared with IL-1α effects (Fig. 1f,h), possibly due to the multiple targets and intracellular pathways engaged by oAβ and its variable aggregation states. Notably, treatment with S3QEL1.2 fully blocked oAβ-induced mtROS (Fig. 1i,j), revealing a pronounced role of CIII in oAβ-induced mtROS responses. Unlike IL-1α and oAβ, stimulation with IL-6 did not increase astrocytic mtROS across various time points (Extended Data Fig. 2i), consistent with its lack of effect on cellular H2O2 efflux (Fig. 1a) and highlighting the stimulus-specificity of mtROS induction.

To further investigate the involvement of CI/CIII in IL-1α- and oAβ-induced ROS production, we generated mice that express alternative oxidase (AOX) selectively in astrocytes. AOX is a mitochondrial enzyme not normally expressed in mammals, but when ectopically expressed, it localizes to the inner-mitochondrial membrane and consumes oxygen through the oxidation of ubiquinol, the substrate required for CIII ROS production (Fig. 2a)32. By shunting ubiquinol, AOX suppresses CIII ROS33 and other ubiquinol-dependent ROS production sites, including CI. Therefore, AOX can phenocopy SELs under contexts where one or more mtROS sites are active. To verify AOX function, we evaluated its ability to alleviate metabolic deficits induced by dysfunction or inhibition of respiratory complexes III and IV, but not complex I6,32,33,34. As expected (Extended Data Fig. 3a), inhibition of electron transport at CIII with myxothiazol alone, or together with inhibition of CI with rotenone, effectively blocked complex IV-dependent mitochondrial respiration in control fl-stop AOX astrocytes (Extended Data Fig. 3b). In contrast, double-transgenic Aldh1l1-AOX astrocytes maintained substantial oxygen consumption in the presence of myxothiazol, which increases ubiquinol for AOX activation, but not in the presence of rotenone, which lowers ubiquinol (Extended Data Fig. 3b). These results indicate that AOX was functional and activated by increased ubiquinol. Notably, AOX expression did not affect basal ATP or mtROS levels, suggesting that AOX does not alter mitochondrial energy production or redox state in astrocytes at baseline (Extended Data Fig. 3c,d).

Fig. 2: NF-κB and NCLX mediate astrocytic mtROS induction.

a, Schematic depicting electron flow to the full reduction of O2 to H2O at complex IV (top) or the partial reduction to superoxide (ROS) at CIII (middle). Ectopically expressed alternative oxidase (AOX) blunts CIII ROS by consuming ubiquinol (QH2) and O2 without producing ROS (bottom). Schematic was created with Biorender.com. b,c, Quantification of mtHyPer7 in Aldh1l1-Cre/fl-stop AOX or control fl-stop AOX astrocytes treated with vehicle, IL-1α (6 h, 3 ng ml−1) or oAβ (3 h, 300 nM). n = 37–161 cells from 2–7 wells, Kruskal–Wallis with Dunn’s test. d, Quantification of NF-κB induction in WT astrocytes treated with vehicle or IL-1α (n = 6 wells, one-way ANOVA with Tukey’s test; representative blots and additional treatment groups included in Extended Data Fig. 4a). e,f, Quantification of mtHyPer7 after IL-1α (6 h) or oAβ (3 h) co-treatment with vehicle, IL-1Ra (1 μg ml−1), TPCA-1 (1 μM), IKK-16 (3 μM) or CGP 37157 (CGP; 10 μM). Inhibitors were added 1 h before and again at the start of stimulation with IL-1α or oAβ. n = 84–352 cells from 5–14 wells, Kruskal–Wallis with Dunn’s test. g,h, Quantification of mtHyPer7 in Slc8b1-f/f or control WT astrocytes transduced with AAV-CMV-Cre and treated with vehicle, IL-1α (6 h) or oAβ (3 h). n = 37–117 cells from 3–6 wells, Kruskal–Wallis with Dunn’s test. i, Representative images and quantification of time-dependent changes in mitochondrial membrane potential after treatment with vehicle, IL-1α, oAβ. The mitochondrial uncoupler BAM15 was used to depolarize ΔΨm and normalize assay responses. n = 3–21 wells, unpaired two-tailed t-test. Scale bars, 50 µm; 25 µm (inset). j, Linear regression correlational analysis of mtHyPer7 versus TMRM changes after treatment with IL-1α or oAβ compared with vehicle. Data are shown as mean ± s.e.m. Some error bars in d and i are not visible because the error range is smaller than the symbol size. Error bars are omitted from j for clarity, but are derived from same corresponding datasets in i. *P < 0.05; **P < 0.01; ***P< 0.001; NS, not significant. Additional statistical details are included in Supplementary Table 6.

Source data

Consistent with our findings that CIII ROS production is triggered by IL-1α and oAβ, Aldh1l1-AOX astrocytes had suppressed mtHyPer7 responses to IL-1α and oAβ as compared with control fl-stop-AOX astrocytes (decreased by 68.2% and 62.3%, respectively) (Fig. 2b,c). To test whether functional AOX is required for these effects on mtROS, we assessed wild-type astrocytes acutely transfected with functional or a catalytically inactive mutant AOX33,35 and confirmed that functional AOX but not mutant AOX effectively suppressed mtROS responses to antimycin A, IL-1α, and oAβ (Extended Data Fig. 3e–g). These results support that CIII is a major driver of astrocytic mtROS responses to disease-related factors.

To identify the upstream molecular mechanisms required for the induction of CIII ROS, we first tested the involvement of nuclear factor-κB (NF-κB), a transcription factor (TF) activated by IL-1R signalling and associated with astrocytic responses to diverse stimuli36,37. Indeed, IL-1α rapidly increased NF-κB phosphorylation levels within 5 min (Fig. 2d), and this increase preceded and was sustained during CIII ROS induction and was not affected by S3QEL1.2 (Fig. 2d and Extended Data Fig. 4a). Notably, inhibition of NF-κB activation with TPCA-1 or IKK-16 or preventing IL-1R activation with IL-1R antagonist protein (IL-1Ra) potently suppressed mtROS induction by IL-1α and oAβ (Fig. 2e,f), indicating that receptor-mediated NF-κB activation is necessary for the induction of CIII ROS. Inhibition of NF-κB in the absence of stimuli had no effect on basal mtROS (Fig. 2e).

To explore the specific mechanisms within mitochondria promoting CIII ROS production, we examined whether IL-1α altered the expression of astrocytic electron transport chain components. However, subunits of complexes I–V were unchanged after stimulation (Extended Data Fig. 4b). Next, we tested the expression and activity of the mitochondrial sodium-calcium exchanger (NCLX), an inner-membrane protein enriched in astrocytes (Extended Data Fig. 4c,d) that regulates mitochondrial ion homeostasis. NCLX has been suggested to promote CIII ROS in peripheral cells under hypoxia38. Although IL-1α did not induce NCLX expression (Extended Data Fig. 4e), blockade of NCLX ion flux with the potent and selective inhibitor CGP 37157 decreased mtROS induction by IL-1α and oAβ (Fig. 2e,f), suggesting that increases in CIII ROS are dependent on NCLX activity. Of note, CGP 37157 had minimal effects on basal mtROS (Fig. 2e), consistent with CIII ROS having minimal roles in maintaining basal state mtROS in astrocytes. In agreement, Cre-dependent knockout of the NCLX-encoding gene in astrocytes derived from homozygous loxP-Slc8b1 (f/f) mice (Extended Data Fig. 4f) prevented IL-1α- and oAβ-induced increases in mtROS (Fig. 2g,h), further indicating that NCLX is essential for astrocytic mtROS responses.

Notably, mitochondrial membrane potential (ΔΨm) is a strong regulator of NCLX activity39. We found ΔΨm was hyperpolarized in IL-1α- or oAβ-stimulated astrocytes, supporting our findings of increased NCLX activity (Fig. 2i). The temporal dynamics of ΔΨm hyperpolarization closely matched the time frames of mtROS induction (Fig. 1), revealing strong positive correlations between ΔΨm and mtROS responses upon IL-1α or oAβ treatment (Fig. 2j). These results are consistent with CIII ROS responses being linked to increases in ΔΨm-dependent NCLX activity.

CIII ROS induce oxidation of diverse but specific target proteins linked to disease

CIII ROS production is proposed to be a major contributor to redox signalling2, but the exact oxidation targets are largely undefined. Previous studies in isolated mitochondria have suggested that different sources of ROS oxidize unique mitochondrial targets40. However, it is not known if CIII or any other site of mtROS production causes oxidation of select targets in intact cells and whether these putative oxidation patterns are affected by physiological stimuli.

To profile disease-relevant CIII-specific oxidation targets across the entire astrocytic proteome, we performed unbiased stoichiometric cysteine-redox proteomics in astrocytes stimulated with IL-1α with or without S3QEL1.2 co-treatment and compared with vehicle as control (Fig. 3a)41. We identified and quantified the percent reversible oxidation for a total of 8,292 unique cysteine-containing peptides (cys peptides), with nearly identical overall ion intensity, cysteine abundance and coverage across the proteome between conditions (Fig. 3b,c and Supplementary Table 1).

Fig. 3: CIII ROS oxidize disease-relevant cysteines across the proteome.

a, Workflow of the stoichiometric redox proteomics method applied to astrocytes 6 h after treatment with vehicle, IL-1α (3 ng ml−1) or IL-1α with S3QEL1.2 (3 µM). n = 6 wells. Schematic adapted from ref. 41 was created with Biorender.com. b, Raw and summed TMT ion intensities averaged across replicates for each condition. c, Raw cys-peptide counts averaged across replicates for each condition and Venn diagram of all cys peptides detected in each condition. d, Volcano plot of 8,292 identified cys peptides across all samples, unpaired two-tailed t-test. Maroon and blue circles highlight individual cysteines whose oxidation increased or decreased by 10% or more after IL-1α treatment relative to vehicle. e, Comparison of baseline oxidation of cys residues susceptible or not susceptible to IL-1α-dependent modification. Unpaired two-tailed t-test: ***P < 0.001. f, Cys peptides grouped according to IL-1α-induced effects. g, Volcano plot showing IL-1α-induced oxidized cys peptides modified by co-treatment with S3QEL1.2, unpaired two-tailed t-test. h, Proportions and plots of IL-1α-induced oxidized cys peptides based on effect of S3QEL1.2. i, Gene Ontology (GO) enrichment analysis for cys peptides oxidized ≥10% by IL-1α and reduced ≥5% by S3QEL1.2 co-treatment. Plot displays the ratio of the number of peptides identified relative to the number of proteins in a GO term. P values were calculated using the cumulative hypergeometric test and the significance threshold was set as P < 0.05 following g:SCS multiple testing correction within g:Profiler. j, Heatmap of the effects of IL-1α with or without S3QEL1.2 on redox states of 183 cys peptides oxidized ≥10% after IL-1α treatment. Specific redox-active cys residues in proteins linked to immune, metabolic and neuropathogenic processes are highlighted. k, Subcellular distribution of redox-active proteins influenced by IL-1α and S3QEL1.2. Additional statistical details are included in Supplementary Table 6.

Source data

Although IL-1α substantially increased astrocytic mtROS levels (Fig. 1), we did not detect global increases in oxidation states in the cysteine proteome. Instead, approximately 4% of mitochondrial cys peptides detected (29 out of 707) and 5% of total cys peptides detected (403 out of 8,292) changed their baseline oxidation by 10% or more at the time of maximal mtROS induction by IL-1α (Fig. 3d; also Fig. 1f), a cutoff consistent with previous reports on meaningful effects of cysteine modifications41,42. Of note, cysteines that were susceptible to IL-1α-induced oxidation had higher baseline oxidation levels (median of 31% oxidized) compared with cysteines in the total proteome (median of 8% oxidized) (Fig. 3e), suggesting that this subpopulation of cysteines in the astrocyte proteome is on average more redox-active at baseline and responsive to cell stimulation, possibly due to their specific intracellular microenvironment and redox properties.

Consistent with CIII ROS promoting oxidative changes, S3QEL1.2 inhibited IL-1α-induced oxidation of the astrocytic proteome. Among the 183 cys peptides (from 170 unique proteins) that were oxidized 10% or more after IL-1α (Fig. 3f,g), 59% were decreased at least 10% by S3QEL and 85% were decreased at least 5% by S3QEL (Fig. 3h and Supplementary Table 1). Proteins susceptible to IL-1α-induced oxidative changes were highly enriched for metabolic processes (Extended Data Fig. 5a). S3QEL reversed many of these changes, particularly pathways and enzymes involved in NADPH-related metabolism (Fig. 3i), supporting a role for CIII ROS in IL-1α-mediated redox signalling. Top CIIIROS-sensitive cysteines included proteins linked to immune pathways (NFKB1, PP6R1, IL6RB and TRAF2), lipid metabolism (FAS, ACSA, ETFD, HMGCL, MECR and CLYBL), dementia-related pathways (GPC4) and redox-regulation pathways (GSHR and AMPL) (Fig. 3j). Our results validate metabolic proteins previously reported to be targets of antimycin A-induced ROS in isolated mitochondria (for example, ETFD and CLYBL)40 and more specifically identify the individual cysteines in these proteins that are sensitive to CIII ROS.

In support of an intracellular signalling role for CIII ROS beyond mitochondria, S3QEL-sensitive cysteines were found across multiple intracellular compartments, suggesting that CIII ROS can diffuse to oxidize extra-mitochondrial targets or engage in long-distance relays to promote cysteine modifications at multiple distinct intracellular sites (Fig. 3k). Overall, these results indicate that CIII ROS are major contributors to increased protein oxidation upon IL-1α stimulation and represent a previously unrecognized source of redox-based selective modifications that include a number of disease-associated proteins. These findings provide comprehensive characterization of wide-ranging protein oxidation across multiple subcellular compartments caused by mtROS and demonstrate that a single mitochondrial site induced by a disease-relevant factor can promote these effects.

Next, we performed polar metabolomic profiling to determine whether IL-1α promotes CIII ROS by altering metabolic pathways or whether CIII ROS are acting upstream of metabolic changes. IL-1α-stimulated astrocytes showed marked trends of increases in specific metabolites (Extended Data Fig. 5b and Supplementary Table 2), including the tricarboxylic acid cycle intermediates malate and fumarate as well as argininosuccinate, which can generate fumarate and participate in the urea cycle (Extended Data Fig. 5c‒e). Inhibition of NF-κB with TPCA-1 lowered many metabolites relative to IL-1α alone (Extended Data Fig. 5f and Supplementary Table 2). Co-treatment with S3QEL1.2 also suppressed the changes (Extended Data Fig. 5g and Supplementary Table 2) and influenced ascorbate-related redox-metabolism and carbohydrate-metabolism pathways (Extended Data Fig. 5h). Integration of S3QEL-sensitive changes in cysteine oxidation and metabolite levels confirmed central carbon metabolism, the tricarboxylic acid (TCA) cycle and antioxidant-related pathways were coordinately regulated by CIII ROS production (Extended Data Fig. 5i,j). These findings indicate that astrocytic CIII ROS influence redox-associated metabolic rewiring upon IL-1α-induced NF-κB signalling.

CIII ROS promote astrocytic STAT3 activity and gene expression linked to neural deficits

Gene expression changes are integral to astrocytic responses in various neurological disorders14,43, but whether mtROS influence astrocytic transcriptional activities is not clear. To investigate if induction of mtROS regulates specific transcriptional programs, we performed untargeted RNA sequencing in IL-1α-stimulated astrocytes in the presence or absence of S3QEL1.2 or S1QEL2.2. As expected26, IL-1α profoundly altered the transcriptional landscape of astrocytes (Fig. 4a and Supplementary Table 3). Notably, CIII ROS suppression with S3QEL strongly altered approx. 30% of IL-1-responsive genes (Fig. 4b) and influenced over tenfold more IL-1-responsive genes than CI ROS suppression with S1QEL (Fig. 4c,d and Supplementary Table 3). S3QEL alone and S1QEL alone each altered only a small number of basal transcripts (Extended Data Fig. 6a,b and Supplementary Table 3). These results indicate that different sites of mtROS influence distinct, wide-ranging transcriptional programs and further highlight the major role of CIII ROS induction in regulating astrocytic neuroimmune responses.

Fig. 4: CIII ROS promote context-specific STAT3 signalling and changes in gene expression.

a–c, Volcano plots of gene expression changes after 6 h treatments with vehicle, IL-1α (3 ng ml−1), or IL-1α with S3QEL1.2 (3 μM) or S1QEL2.2 (10 μM). n = 3–6 wells. P values were calculated using parametric tests and corrected for multiple comparisons using the Benjamini–Hochberg method. d, Comparison of IL-1-responsive genes altered by S3QEL and/or S1QEL. e–g, TF network analysis of gene expression changes after treatment with vehicle, IL-1α, or IL-1α with S3QEL1.2 or S1QEL2.2. n = 3–6 wells, listed factors exceed the P < 0.05 threshold, P values were calculated using Fisher’s exact test within QIAGEN Ingenuity Pathway Analysis (IPA). h–k, Representative western blots and quantification of p-STAT3(Y705) relative to total STAT3 levels after 0.5 h (h) or 6 h (i–k) treatments with vehicle, IL-1α, oAβ (300 nM), S3QEL1.2 or S1QEL2.2. n = 4–13 wells, one-way ANOVA with Tukey’s test. l,m, Representative western blots and quantification of p-STAT3(Y705) levels after 1 h pre-treatment with CGP 37157 (CGP; 10 μM), then 6 h treatment with vehicle, IL-1α,