Three mammalian VDAC isoforms distinctly regulate mitochondrial function and proteome to maintain cell metabolism

Abstract

The voltage-dependent anion channel (VDAC) is the most ubiquitous protein in the mitochondrial outer membrane. This channel facilitates the flux of water-soluble metabolites and ions like calcium across the mitochondrial outer membrane. Beyond this canonical role, VDAC has been implicated, through interactions with protein partners, in several cellular processes such as apoptosis, calcium signaling, and lipid metabolism. There are three VDAC isoforms in mammalian cells, VDAC1, VDAC2, and VDAC3, with varying tissue-specific expression profiles. From a biophysical standpoint, all three isoforms conduct metabolites and ions with similar efficiency. However, isoform knockouts (KOs) in mice lead to distinct phenotypes, which may be due to differences in VDAC isoform interactions with partner proteins. To understand the functional role of each VDAC isoform within a single cell type, we created functional KOs of each isoform in HeLa cells and performed a comparative study of their metabolic activity and proteomics. We found that each isoform KO alters the proteome differently, with VDAC3 KO dramatically downregulating key members of the electron transport chain while shifting the mitochondria into a glutamine-dependent state. Importantly, this unexpected dependence of mitochondrial function on the VDAC3 isoform is not compensated for by the more ubiquitously expressed VDAC1 and VDAC2 isoforms. In contrast, VDAC2 KO did not affect respiration but upregulated electron transport chain components and decreased key enzymes in the glutamine metabolic pathway. VDAC1 KO specifically reduced glycolytic activity linked to decreased hexokinase localization to mitochondria. These results reveal nonredundant roles of VDAC isoforms in cancer cell metabolic adaptability.

Voltage-dependent anion channel (VDAC), a mitochondrial outer membrane (MOM) protein, is crucial for the exchange of small ions such as calcium and water-soluble metabolites such as ATP, ADP, and NADH between the cytosol and mitochondria. This protein forms monomeric weakly anion-selective ß-barrel pores that not only conduct but also effectively regulate metabolite fluxes across MOM. In vitro studies with VDAC reconstituted into the planar lipid membranes showed that one of the mechanisms of such regulation is VDAC's characteristic and evolutionarily conserved voltage-dependent gating. Under applied voltage, the channel moves from a high-conducting unique "open state" permeable for ATP, to low-conducting so-called "closed states" that do not allow passage of ATP (1) but are more permeable to calcium (2). Recent studies in vitro have revealed that metabolite and calcium fluxes through VDAC can also be regulated by VDAC complexation with various cytosolic proteins, such as dimeric tubulin and α-synuclein (3, 4, 5). Furthermore, VDAC is known to be involved in the regulation of calcium signaling through the formation of complexes with calcium transporters on the endoplasmic reticulum (6), sarcoplasmic reticulum (7, 8), and lysosomes (9). VDAC is also implicated in apoptosis through its interaction with several pro- and anti-apoptotic proteins (10, 11, 12, 13, 14, 15, 16). This highlights the intricate role of VDAC in regulating metabolism, calcium signaling, and cell death, which is further complicated by the isoform-specific differences in VDAC interactome and function.

There are three VDAC isoforms in mammals: VDAC1, VDAC2, and VDAC3. VDAC1 and VDAC2 diverged from VDAC3, which is considered the oldest isoform (17, 18). All three isoforms show high sequence similarity (∼70%) and form similar anion-selective voltage-gating channels in vitro based on single-channel electrophysiology experiments (19, 20). However, mice knockout (KO) studies suggest distinct roles for each VDAC isoform, with VDAC2 KO resulting in development delays and embryonic lethality (21, 22). VDAC1 KO mice exhibit mild bioenergetic defects (23) and VDAC3 KO results in male infertility (24). The latter was attributed to each isoform's tissue-specific role and expression level since VDAC3 is the highest expressed isoform in testis. However, VDAC3 KO also results in mitochondrial dysfunction in heart muscles, where it is the least expressed isoform. VDAC1 and VDAC3 demonstrate different affinities for interaction with the two abovementioned VDAC cytosolic regulators—tubulin and α-synuclein (20). Dimeric α/ß-tubulin is the building block of microtubules, and α-synuclein is a neuronal protein implicated in Parkinson's disease. These two abundant cytosolic proteins have no structural, functional, or genetic similarity but possess a common feature—disordered and negatively charged C-terminal tails that act as effective VDAC-blocking domains by being reversibly captured by the VDAC pore and reversing its selectivity to cationic (25). Specifically, VDAC1 binds tubulin and α-synuclein with a ∼100 times higher affinity than VDAC3 (20). In addition, reconstituted VDAC isoforms differ in calcium selectivity, with VDAC3 having a higher calcium preference than VDAC1 (26). De Stefani et al. showed similar results in cells with VDAC3 overexpression, resulting in increased mitochondrial calcium uptake, while VDAC1 may play a role in apoptotic calcium signaling (27). VDAC2 has recently emerged as an important regulator of mitochondrial calcium uptake, especially in the cardiomyocytes (28, 29). Interestingly, although both VDAC1 and VDAC2 are implicated in apoptosis, each of them interacts with a different set of BCL-2 family proteins: VDAC1 preferably interacts with anti-apoptotic BCL2, BCL-X_L (14, 30), and VDAC2 with pro-apoptotic BAK and BAX proteins (21, 22, 31).

VDAC1 and VDAC2 are the most abundant isoforms in most human tissues, and VDAC1 is the only one whose crystal structure has been solved at atomic level of resolution (32, 33, 34). Therefore, rather naturally, the VDAC1 isoform is the most studied in vitro and in vivo, followed by VDAC2. The VDAC3 isoform, in contrast, is the least expressed in vivo (except for testis) and the least biophysically and biochemically studied in vitro. Despite the initial publication by Colombini's and Craigen's labs in 1999 (19), where VDAC3's ability to form typical voltage-gated ion channels has been demonstrated, the following publications often reported that VDAC3 does not form voltage-gated channels (35) or even that it cannot form channels at all (36). Only recently, the record was straightened out in Queralt-Martin et al. work (20), where it was unambiguously established that reconstituted VDAC3, similar to VDAC1 and VDAC2, forms typical anion-selective voltage-gated channels. One of the differences between isoforms is the number of cysteines, with VDAC2 and VDAC3 having seven and nine cysteines, respectively, and VDAC1 having only two cysteines (37, 38). Such a drastic difference focused the studies of VDAC3 on its role in oxidative stress (39), where it was shown that VDAC3 KO in HAP1 cells promoted the accumulation of free radicals dependent on the cysteines, thus suggesting a role for VDAC3 cysteines in countering ROS-induced oxidative stress.

Early studies using mice KO models demonstrated VDAC isoform-specific differences in mitochondrial functions. VDAC1 KO resulted in a universal decrease in the activity of electron transport chain (ETC) complexes, but VDAC3 KO specifically affected mitochondrial function in the heart and sperm (40). VDAC1 KO in mouse embryonic fibroblast cells showed decreased respiration and glycolytic capacity (41). However, until recently, VDAC was mostly ignored in mitochondrial bioenergetics studies as it was historically considered a "molecular sieve" in MOM, except for the studies focused on VDAC1's role in MOM permeabilization at the earlier stages of apoptosis (42, 43) and its tentative involvement in the mitochondrial permeability transition pore (44, 45). Recently, Magri et al. showed that VDAC1 KO diminished complex I–linked respiration in HAP1 cells due to decreased respiratory reserves (46). However, Yang et al. reported no change in respiration but showed increased glycolysis due to VDAC1 KO in H9c2 cells (47). Beyond the mentioned discrepancies, a comprehensive systematic study comparing the role of each VDAC isoform in mitochondrial respiration and metabolism in the same cell type is lacking.

To understand the differences in isoform function in metabolism, we systematically investigated the role of each VDAC isoform KO using CRISPR-Cas9–mediated gene editing in HeLa cells. We found that VDAC1 KO affected glycolysis, while VDAC2 KO did not measurably affect glycolysis or mitochondrial respiration. Surprisingly, it was VDAC3 KO that decreased mitochondrial respiration. Proteomic studies revealed increased expression of mitochondrial proteins upon VDAC2 KO in contrast to extensive downregulation of mitochondrial proteins in VDAC3 KO cells. Overexpression of glutaminases in VDAC3 KO cells, necessary for mitochondrial glutamine metabolism, results in a dependence of these cells on glutamine. We demonstrate that VDAC3, the least expressed isoform, is crucial for mitochondrial function to fuel high-energy demand. Overall, the data reveal nonredundant roles of VDAC isoforms in cancer cell metabolic adaptability.

Results

Characterization of HeLa VDAC isoform KO cells

Functional KOs of each VDAC isoform were generated using CRISPR-Cas9 in HeLa cells. The indels were confirmed by sequencing (Fig. S1), and the KO was confirmed by Western blot (Fig. S2). HeLa cells were chosen as they express the characteristic levels of VDAC isoforms found in most human cell types (48, 49). Although different proteins may respond differently to mass spectrometry (MS) quantitation, considering the sequence similarity among the VDAC isoforms, we postulated that the relative abundances for VDAC isoforms are proportional to their measured copy numbers. Through MS observation, VDAC1 is the most expressed isoform (∼50%), followed by VDAC2 (∼30%), and VDAC3 (∼18%) as the least expressed isoform in HeLa cells (Fig. 1A). The metabolic activity for all VDAC KO clones was investigated by measuring NADH (Fig. S3A), flavin adenine dinucleotide (FAD) (Fig. S3B) fluorescence using flow cytometry, and ATP production rates (Fig. S4) using Seahorse XF Pro Analyzer. While VDAC2 KO did not affect NADH and FAD levels, VDAC1 and VDAC3 KOs showed some variability. VDAC1 KO consistently decreased glycolytic ATP production, but mitochondrial ATP production rate was variable (Fig. S4A). VDAC3 KO results in different glycolytic and mitochondrial ATP production rates (Fig. S4C). Such variability between clones is intriguing and suggests a compensatory mechanism.

Figure 1.

Individual VDAC isoform KO is not compensated by overexpression of another VDAC isoform. A, VDAC isoform expression normalized to VDAC1 expression in WT HeLa cells measured by mass spectrometry. B, bar graphs comparing the changes in the expression of VDAC1, VDAC2, and VDAC3 due to isoform KOs compared to WT. C, Western blot VDAC1, VDAC2, and VDAC3 expression levels in HeLa WT and KO clones. Data from three experiments are shown in the bar graph, and the error bars indicate SD from the mean. The symbols represent data from independent experiments. Significance was tested using one-way ANOVA followed by Dunnett's post hoc test (∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.0001). KO, knockout; VDAC, voltage-dependent anion channel.

For this study, we chose to focus on the clones that significantly reduced NADH and FAD; VDAC1 (G8) and VDAC3 (E5) along with VDAC2 (C9) to understand the isoform-specific role of this MOM protein in metabolism in HeLa cells. We chose VDAC1 (G8) and VDAC3 (E5) clones because even though VDAC1 is the major isoform and VDAC3 is the minor, their effect on metabolic activity is surprisingly the same. Using the MS results, the expression level of each VDAC isoform in each of the KOs could be directly compared (Fig. 1B). MS data confirms complete loss of VDAC1 and VDAC2 expression in the corresponding KOs. However, VDAC3 KO cells show ∼30% expression of VDAC3. This may most likely correspond to the expression of a shorter transcript corresponding to the C-terminal of VDAC3 (Supplementary MS table) since Western blot confirms complete KO of VDAC3 (Fig. 1C). Conventional thinking would assume that the overexpression of other VDAC isoforms might compensate for the loss of one isoform. We found only mild changes in VDAC isoform expression in response to specific KO. VDAC2 expression decreased by ∼20% in VDAC1 KO cells, and VDAC3 expression increased (∼40%) in VDAC2 KO cells, but these changes do not compensate for the KO of each isoform (Fig. 1B). The KO cells do not show appreciable difference in cell morphology (Fig. S5A), but VDAC2 and VDAC3 KOs decrease the growth rate of HeLa cells by 30 to 40% in 72 h (Fig. S5, B and C). Cell proliferation was also measured using the CellTiter 96 AQ_ueous One Solution Cell Proliferation Assay. The assay contains an MTS tetrazolium compound (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt) that is reduced by NADPH or NADH in metabolically active cells into a colored formazan product with absorbance at 490 nm. While this assay is widely used to measure the number of viable cells, the absorbance depends on the metabolic activity of cells. Interestingly, only VDAC3 KO significantly decreases the absorbance (∼40–55%) (Fig. S5D), which suggests that VDAC3 KO specifically affects the metabolic activity in HeLa cells.

To decipher the role of each VDAC isoform in metabolism, we compared the metabolic activity using a Seahorse XF analyzer.

VDAC1 KO affects basal glycolysis in HeLa cells

To examine the role of each isoform in glycolysis, we performed a glycolysis stress test, which measures the changes in extracellular acidification rate (ECAR) due to increased glycolytic activity in cells. Figure 2A shows the ECAR trace in WT and KO cells upon the addition of glucose, oligomycin, and 2-deoxyglucose. Basal glycolysis was calculated by subtracting nonglycolytic acidification after adding 2-deoxyglucose (red box) from the increase in ECAR upon adding glucose (blue box). Only VDAC1 KO decreased basal glycolysis. Surprisingly, VDAC3 KO resulted in complete loss of glycolytic capacity in HeLa cells, even though basal glycolysis was unaffected (Fig. 2C). Glycolytic capacity (Fig. 2A green box) measures the ability of cells to increase glycolysis upon the inhibition of OXPHOS with oligomycin, which may be linked to mitochondrial dysfunction in VDAC3 KO cells.

Figure 2.

VDAC1 KO decreases glycolysis in HeLa cells. A, traces of ECAR upon the addition of glucose, oligomycin, and 2-deoxyglucose (2-DG) for WT (black circle), VDAC1 KO (pink square), VDAC2 KO (teal triangle), and VDAC3 KO (purple diamond) cells. Glycolysis (blue box) and glycolytic capacity (green box) were calculated after subtraction from nonglycolytic acidification (red box). Bar graphs comparing glycolysis (B) and the glycolytic capacity (C) between WT (gray), VDAC1 KO (pink), VDAC2 KO (teal), and VDAC3 KO (purple) HeLa cells. Data from 3 to 4 independent experiments are shown in the bar graph, and the error bars indicate SD from the mean. The symbols represent data from independent experiments. Significance was tested using one-way ANOVA followed by Dunnett's post hoc test (∗p < 0.05). KO, knockout; VDAC, voltage-dependent anion channel.

VDAC3 KO decreases mitochondrial respiration in HeLa cells

The mitochondrial stress test was performed to determine the effect of VDAC isoform KOs on mitochondrial respiration. The oxygen consumption rate (OCR) was measured under basal conditions and after adding oligomycin, FCCP, and rotenone/antimycin-A (Fig. 3A). VDAC1 KO and VDAC2 KO showed no significant change in respiration compared to WT HeLa WT cells (Fig. 3B). However, VDAC3 KO significantly decreased basal and maximal respiration (Fig. 3B). Interestingly, VDAC3 KO resulted in a complete loss of spare respiratory capacity (SRC) (Fig. 3C), calculated by subtracting basal respiration from maximal respiration (50). This suggests that HeLa cells with VDAC3 KO are unable to meet increased energy demand during acute cellular stress. SRC depends on the availability of mitochondrial substrates, suggesting a role for VDAC3, the least expressed isoform, in the uptake of mitochondrial substrates in HeLa cells.

Figure 3.

VDAC3 KO affects mitochondrial respiration in HeLa cells resulting in complete loss of spare respiratory capacity (SRC). A, mitochondrial stress test trace of changes in oxygen consumption rate (OCR) upon the addition of oligomycin, FCCP, and antimycin A/rotenone for WT (black circle), VDAC1 KO (pink square), VDAC2 KO (teal triangle), and VDAC3 KO (purple diamond) cells. Basal (blue box), ATP-linked (yellow box), and maximal (green box) respiration were calculated after subtraction from nonmitochondrial respiration (red box). B, bar graphs comparing basal, ATP-linked, maximal respiration, and SRC (C) in WT (gray), VDAC1 KO (pink), VDAC2 KO (teal), and VDAC3 KO (purple) HeLa cells. Data from three independent experiments are averaged and the error bar indicates SD from the mean. The symbols represent data from independent experiments, and the error bars indicate SD. Significance was tested using one-way ANOVA followed by Dunnett's post hoc test (∗p < 0.05, ∗∗p < 0.01). KO, knockout; VDAC, voltage-dependent anion channel.

VDAC shows isoform-specific changes in cellular proteome

To understand the molecular basis of these striking phenotypes of VDAC isoform KOs at the level of protein expression, label-free proteomics was performed on each of the KO cell lines. Using liquid chromatography coupled with tandem MS, we identified and quantified over 9500 unique proteins across all four cell lines: WT, VDAC1 KO, VDAC2 KO, and VDAC3 KO.

Most proteins (9,168) were identified in all studied cell lines, enabling direct comparison without the need for excessive imputation (Fig. 4A). Expression data were subjected to a median normalization (51). A principal component analysis (PCA) demonstrated tight clustering of each of the sample replicates within each cell type compared to different genetic backgrounds (Fig. 4B). Each cell line was separated into distinct regions of the PCA plot, with the WT cells remaining directly between each VDAC KO. These data indicate that each VDAC isoform KO leads to a unique change in the cellular proteome.

Figure 4.

Total proteomics of VDAC isoform KOs reveal isoform-specific proteome remodeling. A, Venn diagram detailing the number of unique proteins identified in each sample group. B, principal component analysis plot demonstrating the stratification of each VDAC KO sample compared to WT cells along principal components, PC1 and PC2. C, volcano plots of each VDAC isoform KO compared to WT cells (N = 3). Proteins with > 1-fold absolute change and p-value < 0.05 are colored blue for downregulation and red for upregulation compared to the WT. Proteins of interest are labeled. D, representative GO term analysis showing terms significantly downregulated upon each VDAC isoform (blue) or upregulated (red) compared to the WT. E, STRING analysis of glycolysis and gluconeogenesis proteins downregulated in all VDAC isoform KOs compared to WT. KO, knockout; VDAC, voltage-dependent anion channel.

To determine the proteins that were most altered in each KO compared to the WT, differential protein expression analysis was performed. Focusing on those proteins with the most significant and largest magnitude of change between the different KOs, we observed alteration in several proteins hitherto not connected to VDAC in the existing literature. In the VDAC1 KO cells, the highest magnitude loss was of the transcription factor LZTS3 (Fig. 4C). Among the most expressed proteins upon VDAC1 KO was the fatty-acid transporter CD36. In the VDAC2 KO cells, the highest magnitude loss was in the protein PLA2G6, a phospholipase, a gene for which loss of function results in a series of infantile neurodevelopmental disorders and Parkinson's disease. PLA2G6 mutations are known to cause mitochondrial deficits, putatively from alterations in lipid homeostasis (52). We also observed the loss of the metallo-reductase STEAP4, which is associated with mitochondrial dysfunction (53). VDAC2 KO led to overexpression of the apoptosis and inflammasome-related protein PYCARD and the upregulation of the E3 ubiquitin ligase FBOX2 protein. Notably, VDAC2 KO led to a downregulation of BAK1. VDAC2 is known to be essential for the import of the pro-apoptotic BAK into the MOM (21). The downregulation of BAK may indicate a response to the reduced mitochondrial import of BAK in the absence of VDAC2. Naghdi et al. also showed the importance of VDAC2 in mitochondrial recruitment of BAK in HepG2 cells and hepatocytes (54). Interestingly, autophagy receptor p62 was downregulated in HeLa cells with VDAC3 KO. The Western blot further confirmed that VDAC3 KO decreases p62 expression (Fig. S6, A and B). This result suggests either an alteration to autophagy pathways or oxidative stress pathways via KEAP1-Nrf2 due to the loss of VDAC3 (55). VDAC3 KO resulted in the loss of cell death and mitochondrial-related protein CASP9, as well as the metabolic protein PDK4. In VDAC3 KO, we observed upregulation of the spermatogenesis protein SPOCK1, consistent with VDAC3's unique role in this developmental process (24). These data indicate that the majority of pathways altered by each VDAC KO are unique.

To investigate the pathways uniquely perturbed by each isoform, we performed unbiased gene ontology analysis of VDAC KO proteomics data to identify groups of pathways both conserved and altered uniquely by different isoform KOs (Fig. 4D). Among pathways downregulated in all VDAC isoform KOs, the loss of glycolysis and gluconeogenesis-related proteins was observed. Proteins associated with these terms include hexokinase 1 (HK1) (Fig. 4E). HK1 decreased by ∼ 34% in VDAC1 KO, ∼45% in VDAC2 KO, and ∼20% in VDAC3 KO. In contrast, HK2 expression is increased (∼50%) in VDAC3 KO (Fig. S7A). HK1 and HK2 contain mitochondrial targeting sequence (56, 57) and have been suggested to bind VDAC on the MOM (58, 59, 60, 61, 62). Previous studies have shown that the mitochondrial localization of HK can regulate energy metabolism (63, 64). We imaged the colocalization of HK1-GFP (Fig. 5A) and HK2-GFP (Fig. S7B) with mitochondria labeled with Omp25-mCherry in HeLa WT and KO cells to visualize changes in HK localization. Though all VDAC isoform KOs downregulated the expression level of HK1, only VDAC1 KO significantly decreased the localization of HK1 (Fig. 5B) and HK2 (Fig. S7C) to the mitochondria. Other lowered proteins in the glycolytic pathway included aldolases A and C, transaldolase, and GAPDH (Fig. 4E). We also observed a downregulation in proteins related to exocytosis in all KO cells (Fig. 4D). Of pathways altered uniquely in each isoform, VDAC1 KO showed a reduction in proteasome regulation and endoplasmic reticulum protein trafficking. VDAC2 KO resulted in a minor downregulation of proteins related to the actin cytoskeleton, which was upregulated in the VDAC3 KO cells compared to WT, along with focal adhesion (Fig. 4D). VDAC3 KO cells showed the strongest effects on mitochondrially annotated proteins with significant losses of proteins involved in the ETC and oxidative respiration. Following these links to the modulation of mitochondrial pathways, we focused on the alterations to the mitoproteome.

Figure 5.

HK1 mitochondrial localization is dependent on VDAC1 isoform. A, representative images of HK1-GFP (green) colocalization with Omp25-mCherry (red) shown in the merged image (yellow) for HeLa cells with WT, VDAC1 KO, VDAC2 KO, and VDAC3 KO. (Scale bar represents 50 μm). B, colocalization analysis of HK1 with Omp25 measured by Pearson’s correlation coefficient (PCC) shows a significant decrease in HK1 mitochondrial localization in VDAC1 KO. Data from three independent experiments are represented. The symbols represent PCC for each cell, and error bars indicate the SD from the mean. Significance was tested using one-way ANOVA followed by Dunnett's post hoc test (∗∗∗∗p < 0.0001). KO, knockout; VDAC, voltage-dependent anion channel.

VDAC isoform KOs rewire the mitoproteome

To focus on the mitoproteome, we segmented those proteins annotated as mitochondrial in the Mitocarta3 database. All VDAC isoform KOs showed some alteration to mitochondrial proteins, with 26 proteins found downregulated in all VDAC KOs (Fig. 6A). VDAC3 KO cells have 116 unique downregulated mitochondrial proteins, consistent with the GO term analysis. In contrast, VDAC3 KO cells had the least number of unique mitochondrial proteins upregulated at 32, compared to VDAC1 with 84 and VDAC2 with 96. Only 14 proteins were found upregulated across all VDAC KOs. These data support the notion that VDAC3 KO shows severe loss of mitochondrial proteins, whereas VDAC1 and VDAC2 KOs appear to show milder alterations in mitochondrial proteins. The STRING analysis of mitochondrial proteins uniquely altered in each VDAC KO showed that VDAC3 KO cells have a large, downregulated cluster of proteins involved in the ETC (Fig. 6B, red box), as described in the GO term analysis. In contrast, VDAC2 KO cells showed an upregulation of a cluster of ETC-related proteins. These results demonstrate that VDAC1 KO causes the least defined alteration to the mitochondrial proteome in HeLa cells despite VDAC1 being one of the most ubiquitous mitochondrial proteins.

Figure 6.

VDAC isoform KOs lead to countervailing remodeling of the mitochondrial proteome. A, Venn diagram showing mitochondrially annotated proteins found to be downregulated or upregulated in each of the three VDAC isoforms compared to WT. B, STRING analysis of mitochondrial pathways uniquely downregulated or upregulated in each of the VDAC isoform KOs. Proteins involved in the electron transport chain (red box) and glutamine pathway (blue box) are highlighted. KO, knockout; VDAC, voltage-dependent anion channel.

VDAC3 KO leads to defects in the ETC

Figure 7A shows that VDAC3 KO caused a substantial loss in expression across most ETC components, whereas VDAC2 KO mostly upregulated ETC components in the MS results. In contrast, VDAC1 KO cells demonstrated the least perturbation in the ETC proteome. Since we did not purify mitochondria for these experiments, the mitochondrial localization of these proteins was not confirmed. The proteomic results were further validated by western blot of a select number of these hits (Fig. 7, B and C). The proteins in complex I (NDUFA9), complex IV (MT-CO1), and complex V (ATP5A) were found to be consistently downregulated in the VDAC3 KO cells. In contrast, VDAC2 KO showed significant upregulation of ATP5A but downregulation of NDUFA9 and MT-CO1. In addition, the VDAC3 KO caused a slight, but not significant, defect in the expression of TCA enzyme citrate synthase (CS), whereas VDAC2 KO upregulated CS. These results confirm that VDAC3 KO caused the most severe mitochondrial defects, consistent with decreased mitochondrial respiration (Fig. 3).

Figure 7.

VDAC2 and VDAC3 KO display widespread upregulation and downregulation, respectively, of electron transport chain components. A, heatmap of altered proteins associated with the mitochondrial electron transport chain in VDAC isoform KO cell lines compared to WT. B, Western blot of selected ETC components across VDAC isoform KO cell lines. C, densitometry quantification of (B). Data for three repeats were averaged in the bar graphs; error bars indicate the SD from the mean. The symbols represent data from each repeat. Significance was tested using one-way ANOVA followed by Dunnett's post hoc test (∗p < 0.05, ∗∗p < 0.01). ETC, electron transport chain; KO, knockout; VDAC, voltage-dependent anion channel.

VDAC3 KO leads to an increased dependence on glutamine metabolism

The STRING analysis of the mitochondrial proteome (Fig. 6B) across VDAC KO lines showed an upregulation of glutamine pathway proteins in VDAC3 KO cells and a downregulation in VDAC2 KO (Fig. 6B, blue box). To get insight into these surprising results, we investigated the expression of other proteins in the glutamine metabolic pathway across the VDAC KO cells.

We found that VDAC3 KO upregulates glutaminases (GLS1/2) and GLUD1/GPT2 (Fig. 8A). GLS1 is a key mitochondrial enzyme that converts glutamine into glutamate, which GLUD1/GPT2 then converts into α-ketoglutarate to feed the TCA cycle. Western blotting further confirmed the upregulation of GLS1 in VDAC3 KO cells (Fig. 8B). These data indicate that VDAC3 KO cells rewire the metabolic pathways, increasing reliance on glutamine for mitochondrial respiration. Notably, the upregulation ofglutaminase is a hallmark of certain cancers, such as ovarian, breast, and colorectal cancer, resulting in their dependence on glutamine (65). Targeting glutaminase in these cancers has been a popular target of therapeutic intervention (66). To determine whether these alterations have functional effects on the metabolic state of the VDAC KO cells, we performed a mitochondrial fuel flex test using a Seahorse XF analyzer to determine their dependencies on glucose, glutamine, and fatty acid to fuel mitochondrial respiration. In this assay, a series of metabolic inhibitors is used to separate the contribution of each fuel source to respiration. UK5099 blocks the mitochondrial pyruvate carrier to inhibit the glucose oxidation pathway; etomoxir inhibits carnitine palmitoyltransferase 1A, which transports long-chain fatty acids into mitochondria, inhibiting the long-chain fatty acid oxidation pathway; and BPTES inhibits GLS1, which converts glutamine to glutamate, inhibiting the glutamine oxidation pathway (Fig. 8C). The relative fuel dependence can be estimated by measuring the oxygen consumption rate in the presence or absence of the fuel pathway inhibitors. WT HeLa cells had equal dependence on glucose and glutamine with a small contribution from fatty acid oxidation (Fig. 8D). Consistent with glycolysis (Fig. 2) and mitochondrial stress test (Fig. 3), VDAC1 KO cells showed a 37% decrease in glucose dependence compared to WT, while VDAC2 KO cells showed no significant changes in fuel oxidation compared to the WT cells (Fig. 8D). However, VDAC3 KO shifted HeLa cell metabolism towards increased (∼30%) glutamine dependence. While fatty acid dependence also increased (60%), it only contributed ∼20% toward HeLa cell metabolism. The increased glutamine dependence demonstrates that the overexpression of GLS and other glutamine pathway proteins leads to a functional change in metabolic phenotype (Fig. 8D).

Figure 8.

VDAC3 KO leads to increased reliance on glutamine metabolism. A, heatmap of altered proteins associated with glutamine metabolism in VDAC isoform KO cell lines compared to WT. B, Western blot of GLS expression across VDAC isoform KO cell lines and the corresponding densitometry quantification. The bar graph represents the average of three repeats. C, simplified metabolic map of mitochondrial fuel pathways and inhibitors used in Agilent seahorse XF mitochondria fuel flex test (Created on BioRender). D, bar graph comparing the glucose, glutamine, and fatty acid dependences for WT (gray), VDAC1 KO (pink), VDAC2 KO (teal), and VDAC3 KO (purple). Data from three independent experiments are represented. The symbols represent data from independent experiments, and error bars indicate the SD from the mean. Significance was tested using one-way ANOVA followed by Dunnett's post hoc test (∗p < 0.05, ∗∗p < 0.01). KO, knockout; VDAC, voltage-dependent anion channel.

VDAC isoform KOs show distinct effect on mitochondrial morphology

VDAC1 and VDAC3 KOs increased mitochondrial fragmentation compared to WT and VDAC2 KO (Fig. 9A). All isoform KOs decreased mitochondrial volume with VDAC1 KO (∼46%) and VDAC3 KO (∼39%) leading to a large decrease compared to VDAC2 KO (∼17%), but only VDAC1 and VDAC3 KOs decreased mitochondrial surface area, branch length, and junctions (Fig. 9B). To further understand the morphological changes, we imaged mitochondria ultrastructure using electron microscopy. WT HeLa cells showed a variety of mitochondrial morphology which may be linked to their ability to adapt to nutrient availability to promote cell growth. VDAC1 KO decreased cristae density while VDAC3 KO led to increase in mitochondria volume and donut shaped mitochondria (Fig. S8). Interestingly, VDAC2 KO increased the cristae density (Fig. S8) of mitochondria compared to WT.

Figure 9.

VDAC isoform KOs lead to distinct mitochondrial phenotypes. A, representative images of mitochondrial network in HeLa cells WT, VDAC1 KO, VDAC2 KO, and VDAC3 KO labeled with mitobright LT deep red (scale bar represents 10 μm) and inset of zoomed region (scale bar represents 5 μm). B, comparison of mitochondrial volume, surface area, branch length, and junctions between HeLa cells WT, VDAC1 KO, VDAC2 KO, and VDAC3 KO. Data from three independent experiments are represented in the box plots. The symbols represent data for each cell; the borders of the boxes define the 25th and 75th percentiles, with the median displayed as lines in the box and error bars indicating the SD from the mean. Significance was tested using one-way ANOVA followed by Dunnett's post hoc test (∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05). KO, knockout; VDAC, voltage-dependent anion channel.

Discussion

We characterized stable KOs of the three VDAC isoforms in HeLa cells. Knocking out both alleles of each VDAC isoform in an identical cellular background enabled us to compare isoform-specific effects across a battery of different assays.

VDAC1 KO decreased basal glycolysis and decreased mitochondria-bound HK1 and HK2. This is consistent with mice KO studies where VDAC1 KO impaired glucose tolerance and reduced mitochondria-bound HK activity (67). VDAC1 KO in H9C2 cells also decreased HK2 bound to mitochondria, even though it did not affect basal glycolysis (47). The different effects on glycolysis could be due to increased glycolysis in HeLa cells compared to H9C2 cells, which prefer oxidative phosphorylation similar to primary cardiomyocytes (68). HeLa cells also express a 10-fold increase in HK2 compared to HEK-293 cells (69), which has been implicated in the increased glycolysis (Warburg effect) characteristic of cancer cells. In our experiment, VDAC1 KO increased mitochondrial fragmentation but did not affect mitochondrial respiration in HeLa cells in contrast to VDAC1 KO in HAP1 cells, which decreased mitochondrial respiration and reserve capacity by increasing complex-I–linked respiration (46). VDAC1 HK1 complexation has been suggested to prevent mitochondrial fission in HeLa cells (70) and difference in HK expression levels may explain the discrepancy in different cell types. Comparing the effect of VDAC1 KO in two oxidative muscles (ventricle and soleus) and a glycolytic muscle (gastrocnemius) showed opposite effects on MOM permeability, (measured as apparent K_m[ADP]), which is dependent on VDAC (71). The ventricle and gastrocnemius muscles showed an increase in K_m[ADP] compared to the soleus, which had decreased K_m[ADP]. Only the soleus muscle showed a decreased rate of respiration in the presence of maximum ADP (V_max). This suggests that while VDAC1 KO only affects the MOM permeability in the ventricle and gastrocnemius muscles, both mitochondrial membranes are affected by VDAC1 KO in the soleus.

Interestingly, VDAC2 KO did not strongly affect metabolism in our assays despite decreasing cell growth. This could be due to the increased cristae density along with the upregulation of certain ETC complexes in proteomics studies. As cancer cells, HeLa WT cells have a variety of mitochondria ultrastructure, which has been linked to its ability to adapt to the environment and grow (72). The increased cristae density due to VDAC2 KO may reduce reductive biosynthesis limiting cell growth. The decreased cell growth in VDAC2 KO could also be due to changes in apoptosis. Proteomics data showed changes in apoptosis-related proteins, such as downregulation of Bak and upregulation of PYCARD in VDAC2 KO cells. This is consistent with past studies highlighting VDAC2's role in apoptosis through complexation with Bak and BAX (21, 22, 31, 54, 73, 74). Further studies are needed to confirm that the decreased cell growth in VDAC2 KO cells is related to changes in mitochondrial biosynthesis and apoptosis pathways in HeLa cells.

Finally, to our surprise, VDAC3 KO decreased cell growth, increased mitochondrial fragmentation, and showed a significant decrease in metabolic activity measured using MTS assay, confirming the decrease in reducing agents such as NADH and FAD measured by flow cytometry. VDAC3 KO decreased mitochondrial ATP and showed severe defects in mitochondrial respiration. Proteomic studies further confirmed the large-scale downregulation of mitochondrial proteins, such as ETC complexes involved in respiration for VDAC3 KO. We also found the upregulation of glutamine metabolism pathway enzymes in accord with increased glutamine dependence in VDAC3 KO cells.

Our results suggest an evolutionarily conserved metabolite transport function for the oldest VDAC3 isoform, given the decrease in mitochondrial respiration, increased glutamine metabolism, and the complete loss of SRC in VDAC3 KO cells. SRC corresponds to cells' ability to increase mitochondrial respiration during high energy demand. VDAC3 KO in mice results in mitochondrial dysfunction in sperm and heart, two of the most energy-demanding cells. VDAC3 KO–induced male infertility could be expected, given the high expression of VDAC3 in the testis (24). However, VDAC3 is only a minor isoform in the heart, and the expression profile of VDAC isoforms is similar to that of HeLa cells (49). Notably, VDAC3 KO mouse heart muscles showed an increased apparent affinity of mitochondria to ADP (K_m[ADP]). In contrast, gastrocnemius, a mixed glycolytic/oxidative muscle, did not show significant changes in K_m[ADP] (40). This suggests cell type–specific changes in MOM permeability in VDAC3 KO mice.

One of the possible explanations for the different roles of VDAC isoforms in cell metabolism could be their different regulation by cytosolic protein partners such as HK, tubulin, α-synuclein, and BCL2 family proteins. Seminal work by Valdur Saks' group showed that isolated mitochondria have lower K_m[ADP] than permeabilized cells (75). This effect was initially linked to cytoskeletal proteins and subsequently shown to be due to free tubulin blocking VDAC (3). The addition of free tubulin to isolated mitochondria increased the K_m[ADP] of mitochondria, suggesting decreased permeability of MOM to ADP due to VDAC reversible blockage by dimeric tubulin. A second component with unchanged K_m[ADP] was identified, possibly due to a fraction of VDAC remaining open to ADP. This led to the hypothesis that two rates of ADP uptake could be due to differences in the binding affinity of free tubulin to the three VDAC isoforms. Later, Queralt-Martin et al. confirmed in experiments with VDAC reconstituted into planar membranes that tubulin (and α-synuclein) blocks VDAC1 with ∼100 times higher affinity than VDAC3 (20). This suggests that VDAC3 may always be open for metabolite transport in and out of mitochondria, while VDAC1 and VDAC2 are mostly blocked by one of their cytosolic partners, such as tubulin or α-synuclein (76). This is consistent with calculations by Marko Vendelin's group suggesting that only 2% of VDAC channels are open (accessible for cytosolic ADP) in cardiomyocytes (77). Finally, VDAC3 knockdown in HepG2 cancer cells also resulted in the largest decrease in ATP, ADP, and NAD(P)H compared with knockdown of the other two VDAC isoforms (78). These results obtained in HepG2 cells were also interpreted as VDAC1 and VDAC2 being mostly closed by free tubulin and VDAC3 being less sensitive to tubulin. Interestingly, theoretical calculations predicted that VDAC3 may prevent suppression of MOM permeability at elevated MOM potential (79). Overall, these results allow us to speculate that the least expressed VDAC3 isoform is constitutively open in cancer cells for essential metabolite transport function.

VDAC3 is also known to be important for regulating oxidative stress (39), and recent work highlights the protective effect of VDAC3 by allowing metabolite transport in yeast under oxidative stress (80). Oxidative stress has been linked to the regulation of mitochondrial fission-fusion (81). These results highlight the complex regulation of mitochondrial metabolism and morphology by VDAC3 dependent on oxidative stress. Further studies are needed to understand the mechanism of such regulation.

Alternatively, VDAC3 expression may constitute part of a signaling axis that promotes oxidative phosphorylation similar to VDAC2's functions in apoptosis, not through direct channel properties but through the recruitment of Bak. VDAC3 may be a critical signaling intermediate via protein–protein interactions that ultimately promote nuclear expression of pro-OXPHOS gene expression through PGC1-α or an alternative pathway (82). To test this hypothesis, future studies are needed to determine the interacting partners unique to VDAC3, using immunoprecipitation-MS. Additionally, following the approach performed for VDAC2's binding to Bak (74), VDAC2/VDAC3 chimeras can be synthesized to determine which regions of the channel are critical for maintaining the respiration phenotype.

In conclusion, comparing the role of each VDAC isoform in a single cell type allowed us to reveal the distinct roles of each VDAC isoform. Consistent with previous studies, we show that VDAC1 and VDAC2 are involved in glycolysis and apoptosis, respectively. To our surprise, VDAC3, the oldest but minor VDAC isoform, which was initially assumed not to form channels, plays the most crucial role in regulating mitochondrial function and metabolic pathways in HeLa cells. We propose that VDAC3 uniquely enables cells to sustain higher energy demands, explaining its role in energy-demanding cell types such as the heart and sperm. Determining whether VDAC3's effect on metabolism is direct or happens via a yet undescribed cellular signaling pathway represents the next avenue for interrogation. Future studies are needed to understand clonal variability in CRISPR-Cas9 KO cell lines and the role of the VDAC3 isoform in cancer metabolism.

Experimental procedures

Generation of VDAC isoform KO cell lines

CRISPR Cas9–mediated KO cell clone of VDAC1, VDAC2, and VDAC3 in HeLa cells were generated by EditCo Bio, Inc. To generate these cells, ribonucleoproteins containing the Cas9 protein and synthetic chemically modified sgRNA (Table 1) were electroporated into the cells using EditCo's optimized protocol. Editing efficiency was assessed upon recovery, 48 to 72 h post electroporation. Genomic DNA was extracted from a portion of the cells, PCR amplified, and sequenced using next-generation sequencing. The resulting sequencing was analyzed using EditCo's proprietary NGS analysis software. To create monoclonal cell populations, edited cell pools were seeded at one cell/well using a single-cell printer into 96- or 384-well plates. All wells were imaged every 3 days to ensure expansion from a single-cell clone. Clonal populations were screened and identified using the PCR-NGS-analysis genotyping strategy described above.

Table 1.

Guide RNA sequences for CRISPR-Cas9 KO of VDAC isoforms

Gene	Guide RNA sequence
VDAC1	GCAACACUCACCAUAGCCCU
VDAC2	GUCUAUUUGUUGAAGGUUUU
VDAC3	UGUGUGUAUAGGCUUUGGCA

Cell culture

The cells were grown in Dulbecco's Modified Eagle Medium (DMEM, Gibco, 15607) supplemented with 10% fetal bovine serum (Gibco, 10437) at 37 °C and 5% CO₂.

Cell proliferation assay

Cells were seeded at 2500 cells per well in 96-well plates (Thermo Fisher Scientific, 165305), and cell count was measured every 24 h. Cells were washed with PBS and fixed with 4% paraformaldehyde (Electron Microscopy Sciences, 15710) containing 25 μM Hoechst 33342 (Thermo Fisher Scientific, 62249) for 15 min. Cells were washed with PBS and imaged using a DAPI filter in Lionheart FX (BioTek).

CellTiter 96 AQ_ueous One Solution Cell Proliferation Assay (Promega, G3582) was performed based on manufacturer protocol. Cells were seeded at 2500 cells in a 96-well plate and incubated at 37 °C in a 5% CO₂ incubator for 72 h. 20 μl of CellTiter 96 AQ_ueous One Solution Reagent was added to 100 μl media per well and incubated for 1 h at 37 C in a 5% CO₂ incubator. The absorbance was measured at 490 nm using a CLARIOstar plate reader.

Seahorse metabolic flux assay

Cells were seeded at 10,000 to 15,000 cells/well in Seahorse XFp PDL miniplates (Agilent, 103721) or XFe96/XF Pro PDL cell culture microplates (Agilent, 103798). They were incubated overnight at 37 °C and 5% CO₂. The following day, the media was replaced with Seahorse XF DMEM assay media (Agilent, 103575) supplemented with glucose, pyruvate, and glutamine based on the manufacturer's protocol for cell mitochondrial stress (Agilent, 103010), glycolysis stress (Agilent, 103020), and mitochondria fuel flex (Agilent, 103260) assays. The assays were performed on Seahorse XF HS mini or XF Pro analyzers (Agilent) according to the manufacturer's protocol. Hoechst 33342 (Thermo Fisher Scientific, 62249) was added to the last port to stain nuclei for cell count and imaged using Lionheart FX or Cytation5 (BioTek).

Colocalization of HK to mitochondria

Cells were seeded at 25,000 cells/well in μ-Slide 8 well chambered coverslip (Ibidi, 80807) and grown at 37 °C and 5% CO₂ overnight. Cells were transfected with HK1-GFP (Addgene, 21917) or HK2-GFP (Addgene, 21920) along with Omp25-mCherry (Addgene, 157758) using Lipofectamine Stem transfection reagent (Thermo Fisher Scientific, STEM00008) according to the manufacturer's protocol. The cells were washed and imaged in Fluorobrite DMEM (Gibco, A18967) with a 20x/0.75 N. A. objective in Leica TCS SP8 microscope. GFP was excited with a 488 nm laser and emission was captured using a 500 to 582 nm filter, and mCherry was excited with a 587 nm laser and emission was captured using a 592 to 784 nm filter. Images were optimized for contrast and brightness, and Pearson's correlation coefficient was measured using the BIOP JACoP plugin in ImageJ (NIH).

Mitochondrial morphology fluorescence imaging

Cells were seeded in a μ-slide 15 well 3D glass bottom coverslip (Ibidi, 81507) and grown at 37 °C and 5% CO₂ overnight. The cells were treated with 10 nM of mitobright LT deep red (Dojindo, MT12) for 1 h at 37 °C in a 5% CO₂ incubator. The cells were washed with prewarmed DMEM three times and fixed with 2% glutaraldehyde (Sigma-Aldrich, G5882) for 15 min at room temperature. The fixation reaction was quenched with 0.1 M NH₄Cl for 5 min at room temperature and then washed with PBS five times before adding mounting media (Abberior, MM-2013). After 24 h, the cells were imaged using Leica TCS SP8 microscope with a 100x/1.4 N. A. oil objective and excited at 633 nm and emission captured from 643 to 784 nm. 3D z-stacks were quantified using ImageJ plugin Mitochondria analyzer (83).

Electron microscopy

Cells were plated on Aclar 33C plastic coverslips (EMS, 50425). Coverslips were prepared by cutting the Aclar 33C film into approximately 12 mm squares to fit in the wells of a 12-well culture plate. Coverslips were cleaned and sterilized by rinsing 10 times in distilled water, followed by 70% ethanol, and finally sterile distilled water. Prior to plating cells, coverslips were rinsed once with DMEM (Gibco, 15607) supplemented with 10% fetal bovine serum (Gibco, 10437) and then cells were plated and grown at 37 °C and 5% CO₂. When cells reached about 75% confluence, they were fixed with 2% glutaraldehyde, 2% paraformaldehyde, and 2 mM calcium chloride in 0.1 M cacodylate buffer (CB) (EMS, 11652) for at least an hour at room temperature. Cells were then rinsed three times with CB and stained with reduced osmium containing 0.25% osmium tetroxide (EMS, 19150) and 0.25% potassium ferrocyanide (Sigma-Aldrich, 3289) in CB for 40 min. Next, cells were rinsed three times with CB and incubated in 1% tannic acid (EMS, 21700) in CB for 30 min. After three rinses with CB, cells were rinsed with 50 mM sodium acetate buffer (EMS, 21120), pH 5.2, and then incubated in 1% uranyl acetate (EMS, 22400) in acetate buffer for 30 min. Cells were then rinsed with acetate buffer followed by dehydration through a graded series of 5-min ethanol rinses (50%, 75%, 95%, 100% twice) prior to infiltration in EmBED812 epoxy resin using the manufacturer’s hard resin formulation (EMS, 14120). To embed, resin-filled gelatin capsules (EMS, 70110) were inverted and placed over the cells on the coverslip and polymerized at 60 C for 2 days. Ultrathin sections of the cells were cut to a thickness of 70 nm using an ultramicrotome (Leica Microsystems Inc.) with a diamond knife (DiATOME) and collected on 1 mm-slot grids with a formvar support film (EMS, Cat. No. FF2010-CU). Sections were viewed using a JEOL1400 Flash transmission electron microscope (JEOL USA, Inc.) operated at 120 KV, and images were recorded with a Biosprint29 CMOS detector (Advanced Microscopy Techniques) or a Tecnai T20 electron microscope (Thermo Fisher Scientific) operated at 200 KV, and images were recorded with a NanoSprint1200 CMOS detector (Advanced Microscopy Techniques).

Western blot

After harvesting, cell pellets were resuspended in 150 μl lysis buffer (50 mM Tris–HCl pH 8.0, 150 mM NaCl, 1% Triton X-100 with protease inhibitor tablet (Pierce)) and incubated on ice for 10 min with occasional vortex. Samples were centrifuged at 15,000 rpm at 4 °C for 10 min, and 120 μl of the supernatant was transferred into a new 1.5 ml tube. Total soluble protein concentrations were measured using the Bradford reagent (Bio-Rad, 5000006). After that, 40 μl of 4x Laemmli sample buffer (Bio-Rad, 161–0774) containing 0.1 M DTT (Cytiva, 17-1318-02) was mixed with the samples and heated for 5 min at 95 °C. Equal amounts of protein samples were loaded and separated using 4 to 20% Mini-PROTEAN TGX precast gels (Bio-Rad, 456–1096) and transferred to nitrocellulose membranes using Trans-Blot Turbo Transfer System (Bio-Rad, 170–4155). The membrane was incubated with Ponceau-S stain for 5 min and imaged. The membrane was destained with Tris-buffered saline-Tween (TBST) buffer and blocked with 5% w/v nonfat milk prepared in TBST buffer, incubated with primary antibodies (Table 2) overnight at 4 °C, washed with 5% milk (three times for 10 min each), incubated with proper secondary antibodies for 2 hours at room temperature, and washed with TBST buffer (three times for 5 min each). The blots were imaged using an ECL reagent (MilliporeSigma, WBKLS0500) and the ChemiDoc MP Imaging System (Bio-Rad). Blot densities were analyzed using Image Lab 6.0.1 software (Bio-Rad) and normalized to total protein density from Ponceau-S staining as loading control.

Table 2.

Primary antibodies for western blot

Antibodies	Source	Identifier
Mouse Anti-MT-CO1	Abcam	ab14705
Mouse Anti-NDUFA9	Abcam	ab14713
Mouse Anti-NDUFB7	Santa Cruz	Sc-365552
Rabbit Anti-GLS1	Proteintech Group	29519-1-AP
Rabbit Anti-CS	Proteintech Group	16131-1-AP
Rabbit Anti-ATP5α	Abcam	ab176569

Label-free proteomics

1.2 million cells were harvested with trypsin, quenched with PBS, spun down to remove the trypsin, and resuspended and spun down in ice-cold PBS. The supernatant was removed, and LC-MS samples were prepared using the Thermo EasyPep Mini MS Sample Prep Kit (Thermo Fisher Scientific, A4006) according to the manufacturer's instructions. Samples were then resuspended in 0.1% formic acid (Thermo Fisher Scientific, 85178) solution, and peptide concentration was tested using the Pierce Quantitative Fluorometric Peptide Assay (Thermo Fisher Scientific, 23290). LC-MS/MS experiments were performed by loading a 500 μg sample onto an EASY-nLC 1000 (Thermo Fisher Scientific) connected to an Orbitrap Eclipse Tribrid mass spectrometer (Thermo Fisher Scientific). Peptides were separated on an Aurora UHPLC Column (25 cm × 75 μm, 1.6 μm C18, AUR2-25075C18A, Ion Opticks) with a flow rate of 0.4 ml/min and for a total duration of 131 min. The gradient was composed of 3% solvent B for 1 min, 3 to 19% B for 72 min, 19 to 29% B for 28 min, 29 to 41% B for 20 min, 41 to 95% B for 3 min, and 95 to 98% B for 7 min. Solvent A consisted of 97.8% H₂O, 2% acetonitrile, and 0.2% formic acid, and solvent B of 19.8% H₂O, 80% acetonitrile, and 0.2% formic acid. MS1 scans were acquired with a range of 400 to 1600 m/z in the Orbitrap at 120 k resolution. The maximum injection time was 50 ms, and the automatric gain control target was 23105. MS2 scans were acquired using quadrupole isolation mode and higher-energy collisional dissociation activation type in the Iontrap. The isolation window was 1.4 m/z, collision energy was 35%, maximum injection time was 35 ms, and the automatric gain control target was 13104. Other global settings were set to the following: ion source type, NSI; spray voltage, 2500 V; ion transfer tube temperature, 300 °C. Method modification and data collection were performed using Xcalibur software (Thermo Fisher Scientific).

Quantification and statistical analysis

Proteomic analysis was performed using Proteome Discoverer 2.4 (PD 2.4, Thermo Fisher Scientific) software, the Uniprot human database, and SequestHT with Percolator validation. Percolator false discovery rates were set at 0.001 (strict) and 0.01 (relaxed). Peptide false discovery rates was set at 0.001 (strict) and 0.01 (relaxed), with medium confidence and a minimum peptide length of 6. Carbamidomethyl (C) was set as a static modification; oxidation (M) was set as a dynamic modification; acetyl (protein N-term), Met-loss (Protein N-term M), and Met-loss + acetyl (Protein N-term M) were set as dynamic N-Terminal modifications. Protein abundance normalization was performed relative to the total peptide amount. Differential expression analysis was performed with media using a custom Python module following the user guide 7 or one-tailed Student's t test with PD 2.4. Principal component analysis analyses were conducted with custom Python code. Volcano plots were generated in R using the Tidy Proteomics and Enhanced Volcano plot packages. Venn diagram was plotted using Python3. Gene ontology analysis was performed using g:Profiler (website (https://biit.cs.ut.ee/gprofiler/gost). Other statistical analyses were carried out by one-tailed Student's t test or one-way ANOVA using Prism 10.0. Three independent biological replicates were used. p values < 0.05 are reported as statistically significant and are depicted as follows throughout the manuscript: ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

Data availability

The raw and processed proteomics data underlying Figures 1A, B, 4, 6, 7A and 8A are accessible from MassIVE database under the identifier MSV000097150. The authors can provide any additional information required to reanalyze the data reported in this paper upon request.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflicts of interests with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Qi-Qun Tang