Mitochondrial membrane hyperpolarization modulates nuclear DNA methylation and gene expression through phospholipid remodeling

Abstract

Maintenance of the mitochondrial inner membrane potential (ΔΨm) is critical for many aspects of mitochondrial function. While ΔΨm loss and its consequences are well studied, little is known about the effects of mitochondrial hyperpolarization. In this study, we used cells deleted of ATP5IF1 (IF1), a natural inhibitor of the hydrolytic activity of the ATP synthase, as a genetic model of increased resting ΔΨm. We found that the nuclear DNA hypermethylates when the ΔΨm is chronically high, regulating the transcription of mitochondrial, carbohydrate and lipid genes. These effects can be reversed by decreasing the ΔΨm and recapitulated in wild-type (WT) cells exposed to environmental chemicals that cause hyperpolarization. Surprisingly, phospholipid changes, but not redox or metabolic alterations, linked the ΔΨm to the epigenome. Sorted hyperpolarized WT and ovarian cancer cells naturally depleted of IF1 also showed phospholipid remodeling, indicating this as an adaptation to mitochondrial hyperpolarization. These data provide a new framework for how mitochondria can impact epigenetics and cellular biology to influence health outcomes, including through chemical exposures and in disease states.

Fig. 1. Ablation of IF1 increases resting ΔΨm via ATP synthase hydrolytic activity.

A ΔΨm was measured in intact cells using TMRE; data were normalized to mitochondrial content using the ΔΨm-insensitive dye MitoTracker® Green (MTG). On y axis, TMRE/MTG ratio. Data are presented as mean values ± SEM of n = 3 biological replicates. Statistical difference by two-sided unpaired Student’s t test. B Representative trace of concomitant ΔΨm and Ca²⁺ clearance measurements in permeabilized cells with succinate; n = 4 independent biological replicates. The black line represents WT cells, and the red line represents IF1-KO cells. FCCP was added at the end of the experiment to fully depolarize the ΔΨm and release the intramitochondrial pool of Ca²⁺. C Representative immunoblots of blue native (BN)-PAGE. Left panel, digitonin-permeabilized WT (lanes 1, 2) and IF1-KO (lanes 3, 4) mitochondria stained with Coomassie Blue. The predicted size of ATP synthase is pointed out by the arrow. Right upper panel, immunoblots of WT and IF1-KO mitochondria blotted for ATP synthase F_o rotor (c-ring). Right lower panel, WT and IF1-KO mitochondria blotted for IF1. Results were reproduced independently three times and with additional antibodies probing other ATP synthase subunits (Supplementary Fig. 1). D Coomassie-stained BN-PAGE showing ATP synthase. Samples were evaluated for ATP hydrolase activity using in-gel activity assays (upper right gel). Quantification of the hydrolase activity is shown in the bar graphs. Data are presented as mean values ± SEM of n = 4 biological replicates. Statistical difference by two-sided unpaired Student’s t test. E Bar graph of ΔΨm as measured in A in cells grown in glucose (dotted lines) or 5 mM galactose (filled bars). Y axis depicts the TMRE/MTG ratio. Numbers beside each bar represent the % difference (Δ) of the ΔΨm when each cell was grown in high glucose vs. in galactose. Data are presented as mean values ± SEM of n = 3 biological replicates per cell type/condition. Statistical difference by two-sided unpaired Student’s t test.

Fig. 2. IF1-KO cells with increased ΔΨm engage in a transcriptional feedback loop repressing a mitochondrial gene expression program.

A Heatmap of differentially expressed genes (DEGs) in IF1-KO vs. WT (False Discovery Rate, FDR < 0.05). The color scheme is presented as a log2 fold change (log2FC). Shades of red represent upregulated genes and shades of blue represent downregulated genes. B Gene Ontology (GO) analysis of DEGs in IF1-KO vs. WT cells. −Log10 FDR values on the y axis, and combined score on the x axis. In bigger blue circles, significantly enriched pathways (FDR < 0.05) related to mitochondria. In smaller translucid blue and gray circles, other significantly enriched pathways (FDR < 0.05) and non-significant pathways (FDR > 0.05), respectively. C Structural depiction of individual complexes of the electron transport chain and mitochondrial ribosomes. The number of DEGs associated with these complexes and their transcriptional directionality are also depicted. Created in BioRender. Mori, M. (2025) https://BioRender.com/wfvugau. D Functional Annotation Clustering of common DEGs reciprocally inverse changed between IF1-KO vs. WT and IF1-OE vs. IF1-KO. Keywords and terms found in each cluster are shown on the y axis, and the Enrichment score is on the x axis. E Venn diagrams show the intersection between reciprocally regulated DEGs in IF1-KO vs. WT and IF1-OE vs IF1-KO. Upper circles: upregulated DEGs based on IF1-KO vs. WT comparison, lower circles represent the downregulated counterparts. F, G GO analysis of reciprocally regulated DEGs in IF1-KO cells, as described in (B). H Heatmap of representative pathways reversed in IF1-OE vs. IF1-KO as identified by GO analysis. Color scheme as described in (A). KO is IF1-KO vs. WT, and OE is IF1-OE vs. IF1-KO.

Fig. 3. Promoters of mitochondrial and phospholipid genes are hypermethylated in IF1-KO cells.

A Heatmap of differentially methylated and expressed genes (DMEGs); left track data from RNA-seq and right track data from DNA methylation (5meC). Data are presented as a log2 fold change (log2FC) for gene expression, the average change in DNA methylation is shown as average delta beta (AVGΔbeta). In red: upregulated genes, and downregulated in blue; hypermethylated gene promoters are shown in brown and hypomethylated in purple. DMEGs were grouped in 4 clusters: 1) hypermethylated and upregulated, 2) hypermethylated and downregulated, 3) hypomethylated and upregulated, and 4) hypomethylated and downregulated. The dashed lines highlight DMEGs from clusters 2 and 3. B, C GO analysis of DMEGs of clusters 2 and 3, respectively. −Log10 FDR values on the y axis and combined score on the x axis. D, E Bar graph of ΔΨm as measured by TMRE normalized to MTG. Y axis depicts the TMRE/MTG ratio. Data are presented as mean values ± SEM of n = 3 biological replicates. Statistical difference by two-sided one-way ANOVA with Tukey’s post-test. F Violin plots showing average probe signal in the transcriptional start site (TSS) across the genome in the four genotypes; the y axis depicts average probe values; white lines indicate the median and dotted lines the quartiles. Numerical values below the violin plot are the mean DNA methylation level of TSSs for each genotype (n = 4). Significance based on on two-sided one-way ANOVA with Dunnett’s correction ****p = <0.0001. G Heatmap of locus-specific DNA methylation from clusters 2 and 3 genes (from A); each line represents the same locus. The color scheme is presented as AVGΔbeta. Hypermethylated gene promoters in brown, hypomethylated in purple. Left track (KO) IF1-KO vs. WT, middle track (OE) IF1-OE vs. IF1-KO, and right track (UCP4) IF1-KO UCP4 vs. IF1-KO (n = 4). H Heatmap of representative pathways as in Fig. 2H (n = 3). I Heatmap of biological processes of mitochondrial genes from DMEGs of each paired comparison. The color scheme is presented as FDR. Comparisons are: left track (KO) IF1-KO vs. WT, middle track (OE) IF1-OE vs. IF1-KO, and right track (UCP4) IF1-KO UCP4 vs. IF1-KO.

Fig. 4. Neither metabolites or redox changes seem associated with epigenetic changes upon mitochondria hyperpolarization.

A Measurement of metabolites involved in DNA (de)methylation reactions per whole cell steady-state metabolomics. Mean WT was set as 1 (n = 6 per genotype; error bars represent ±SEM). On the y axis, fold changes (FC) of metabolites relative to WT (black), red represents the IF1-KO and blue the IF1-OE counterparts. B Joint-pathway analysis using genes and metabolites differently enriched as per RNA-seq and metabolomics analyses, respectively. −Log10 FDR values on the y axis, and combined score on the x axis. Highlighted in black circles are the pathways of interest. C Oxidized/reduced glutathione (GSSG/GSH) ratio based on metabolomics data. On the y axis, GSSG/GSH ratio relative to WT. Data are presented as mean values ± SEM of n = 6 biological replicates. Statistical difference by two-sided one-way ANOVA with Tukey’s post-test. D GSSG/GSH as per degree of Grx1-roGFP probe oxidation in the cytosol (cyto, construct is targeted to the cytosol) (n = 4) and mitochondria (mito, probe is targeted to the mitochondria) (n = 3). Data are presented as mean values ± SEM. Statistical difference by two-sided unpaired Student’s t test. E Total NADP(H) content was estimated using a commercially available kit. Left graph, on the y axis, nmol NADPH or NADP⁺ per μg protein. WT black fill and outline, IF1-KO red fill and outline, IF1-OE blue fill and outline. Filled and outlined bars represent NAPDH and NADP⁺, respectively. Right graph, NADP⁺/NADPH ratio. Data are presented as mean values ± SEM of n = 3 biological replicates with two technical replicates. Statistical difference by two-sided one-way ANOVA with Tukey’s post-test. F H₂O₂ release in the medium as accessed by Amplex® Red. On the y axis, pmol of H₂O₂/min/10⁶ cells. Data are presented as mean values ± SEM of n = 3 biological replicates. Statistical difference by two-sided unpaired Student’s t test.

Fig. 5. Methyl groups are re-routed from phospholipid to DNA methylation when mitochondria are hyperpolarized.

A, B Total (whole cell) and mitochondrial (mito) PC and PE content were analyzed by HPTLC. On y axis, PC and PE in μg/mg of protein. C PC/PE ratio from data in (A, B). On y axis, absolute PC/PE ratio. Data are presented as mean values ± SEM of n = 4 biological replicates (whole cell), and n = 6 biological replicates (mito). Statistical difference by two-sided one-way ANOVA with Tukey post-test. D Schematic diagram for tracing methyl group using deuterated C5-methionine. Carbon (C) is depicted in white circles, nitrogen (N) in blue, sulfur (S) in yellow, and deuterated methyl group (CD₃) in red. Deuterated SAM (CD₃-SAM)is formed from deuterated C5-methionine. Methyl transfer from CD₃-SAM to: 1) PE to form PC yields +3, +6, or +9 isotopologues; and 2) C to form 5meC yields +3 isotopologue. Created in BioRender. Mori, M. (2025) https://BioRender.com/36yccol. E Measurement of PC(16:0/16:0) isotopologues; mean WT was set as 1. On the y axis, fold changes (FC) of PC relative to WT. On the x axis, isotopologues +0, +3, and +6. Black bars, WT; red bar, IF1-KO. F Total 5meC was set as 100% for each sample, and isotopologues +0 and +3 were calculated as % from total. On the y axis, relative abundance of each isotopologue. On the x axis, isotopologues +0 and +3. Black bars, WT; red bar, IF1-KO. Data are presented as mean values ± SEM of n = 3 biological replicates. Statistical difference by two-sided unpaired Student’s t test. G Average probe signal at the TSS of genes in IF1-KO (red) relative to IF1-KO PEMT-OE (violet). Statistical difference by two-sided unpaired Student’s t test (n = 4) **p = 0.0028. H Total PC content in WT (black), WT PEMT-OE (yellow), IF1-KO (red), and IF1-KO PEMT-OE (violet). Data are presented as mean ± SEM of n = 4 biological replicates. Statistical difference by two-sided one-way ANOVA with Tukey’s post-test. I Seahorse flux analyzer was used to estimate oxygen consumption in the cells used in (H). n = 3/genotype. O is oligomycin, F is FCCP, and R/A refer to rotenone/antimycin. J Graphical representation of oxygen consumption rates-derived parameters from I.

Fig. 6. Chronic exposure of WT cells to chemicals that hyperpolarize mitochondria leads to loss in PC, decreased PC/PE ratio, and nuclear DNA hypermethylation.

A Hill slope was used to determine EC50 of chemicals tested. On the y axis, the TMRE/MTG ratio, and on the x axis chemical concentration in log10 μM. Error bars represent ±SD. B Chemical-induced ΔΨm hyperpolarization in cells treated for 10 days. Statistical difference by two-sided one-way ANOVA with Dunnett’s post-test compared to CTL; error bars represent ±SEM of n = 4 biological replicates. C Oxygen consumption rate (OCR) as per the Seahorse Flux analyzer; on the y axis in pmol O₂/min/μg protein. Right graph shows extracellular acidification rate (ECAR) on the y axis in mpH/min/μg protein. O is oligomycin, F is FCCP, and R/A refer to rotenone/antimycin (n = 3); error bars represent ±SEM. D Graphical representation of oxygen consumption rates-derived parameters. Statistical difference by two-sided one-way ANOVA with Dunnett’s post-test compared to CTL; error bars represent ±SEM of n = 3 biological replicates with 6 or 8 experimental replicates based on (C). E, F PC and PE content, respectively, in cells chronically treated with ΔΨm hyperpolarizing chemicals. On the y axis, PC or PE content in nmol/10⁶ cells. G Absolute PC/PE ratio from (E, F) in cells chronically treated with ΔΨm hyperpolarizing chemicals. CTL (black), telmisartan 2 μM (yellow) and annatto 10 μM (green) for (E–G). Statistical difference by two-sided One-way ANOVA with Dunnett’s post-test correction; error bars represent ±SEM of n = 3 biological replicates with 2 experimental replicates for (E–G). H Violin plots showing average TSS probe signal across the genome of cells treated with telmisartan or annatto relative to the vehicle-only-treated WT. White lines indicate the median and dotted lines the quartiles. Numerical values below the violin plot represent the mean methylation level of all TSSs for each treatment; n = 5 biological replicates, statistical analysis done using two-sided one-way ANOVA with Dunnett’s correction. ****p = <0.0001.

Fig. 7. Naturally hyperpolarized cells remodel PC/PE content.

A Fluorescence-activated cell sorting gating parameters. Upper panel: 1st gate for MTG-positive cells with ±10% of the mean, resulting in 30% of the all MTG-positive population. Lower panel: 2nd gate for TMRE-positive cells with lower-, mid-, and upper 5% of TMRE-positive population. Y axis, cell count (events), x axis MTG or TMRE fluorescence in log scale. B PC content in the low-, mid- and upper 5% sorted cells. On the y axis, PC content in fmol/cell. Statistical difference by two-sided one-way ANOVA with Dunnett’s post-test compared to mid; error bars represent ±SEM of n = 3 biological replicates with 2 experimental replicates. C Left graph shows ECAR, on the y axis in mpH/min/μg protein; right graph shows OCR, on the y axis in pmol O₂/min/μg protein (n = 3 for low and mid, n = 2 for high); error bars represent ±SEM. D Graphical representation of OCR-derived parameters from (C). Statistical difference by two-sided one-way ANOVA with Dunnett’s post-test compared to mid; error bars represent ±SEM. E) Representative immunoblots of lysates from Caov3 and OVCAR3 cell lines, WT and IF1-KO cells were used as controls. F ΔΨm measured in intact Caov3 and OVCAR3 cells; y axis, TMRE/MTG ratio. Statistical difference by two-sided unpaired Student’s t test; error bars represent ±SEM of n = 8 biological replicates. G Absolute PC/PE ratio in ovarian cancer cells using lipidomics. Statistical difference by two-sided unpaired Student’s t test; error bars represent ±SEM of n = 3 biological replicates. H Working model: chronic ΔΨm hyperpolarization leads to mitochondrial phospholipid remodeling by decreasing methylation of PC from PE, impacting the cellular PC/PE ratio and leading to a relative increase in PE. This helps regulate proton flux and maintain an optimal hyperpolarized state that is supportive of mitochondrial function. Methyl groups from SAM not used to generate PC are re-routed to the nucleus, rewiring DNA methylation. Differentially methylated loci drive an adaptive gene expression program, including inhibition of OXPHOS that further regulates proton flux and the ΔΨm, redox homeostasis, and glucose/lipid metabolism. Created in BioRender. Mori, M. (2025) https://BioRender.com/96n70n2.