Systems-level analyses dissociate genetic regulators of reactive oxygen species and energy production

Abstract

Respiratory chain dysfunction can decrease ATP and increase reactive oxygen species (ROS) levels. Despite the importance of these metabolic parameters to a wide range of cellular functions and disease, we lack an integrated understanding of how they are differentially regulated. To address this question, we adapted a CRISPRi- and FACS-based platform to compare the effects of respiratory gene knockdown on ROS to their effects on ATP. Focusing on genes whose knockdown is known to decrease mitochondria-derived ATP, we showed that knockdown of genes in specific respiratory chain complexes (I, III, and CoQ10 biosynthesis) increased ROS, whereas knockdown of other low ATP hits either had no impact (mitochondrial ribosomal proteins) or actually decreased ROS (complex IV). Moreover, although shifting metabolic conditions profoundly altered mitochondria-derived ATP levels, it had little impact on mitochondrial or cytosolic ROS. In addition, knockdown of a subset of complex I subunits-including NDUFA8, NDUFB4, and NDUFS8-decreased complex I activity, mitochondria-derived ATP, and supercomplex level, but knockdown of these genes had differential effects on ROS. Conversely, we found an essential role for ether lipids in the dynamic regulation of mitochondrial ROS levels independent of ATP. Thus, our results identify specific metabolic regulators of cellular ATP and ROS balance that may help dissect the roles of these processes in disease and identify therapeutic strategies to independently target energy failure and oxidative stress.

Keywords: ATP; CRISPRi; ROS; metabolism; mitochondria.

Fig. 1.

CRISPR screen for genetic regulators of ROS finds robust, antioxidant-sensitive ROS phenotypes. (A) Schematic describing a mini-library of CRISPRi sgRNAs that identifies ROS phenotypes based on MitoSOX or DCFDA levels via FACS. (B) K562 cells expressing a CRISPRi mini-library and incubated in respiratory conditions for 1 h prior to cell sorting on MitoSOX or DCFDA staining to measure mitochondrial or cytosolic ROS phenotypes, respectively. Several knockdowns associated with complex I (TMEM261, NDUFA8, GRSF1, and NDUFAF1) and CoQ10 biosynthesis (PDSS1, PDSS2, COQ5, and COQ2) robustly increase mitochondrial or cytosolic ROS. Complex IV–associated gene knockdowns (COX18, COX16, and COX11) robustly decrease mitochondrial ROS. (C) High mitochondrial and cytosolic ROS phenotypes associated with complex I knockdown are lowered with antioxidant treatments, either 10 µM MitoQ or 1 mM Trolox. Data compiled from n = 2 experiments. (D) ROS phenotypes exhibited high correlation between metabolic substrate conditions. Data was compiled from n = 2 replicates for basal metabolic conditions, and from n = 3 replicates for respiration-only and glycolytic-only conditions.

Fig. 2.

Systems-level analysis of ATP and ROS phenotypes. (A) Mitochondrial ROS level assessed in K562 cells expressing an expanded library of CRISPRi sgRNAs with MitoSOX or (B) cytosolic ROS with DCFDA were plotted against previously determined ATP phenotypes (13). Knockdown of complex I subunits in the respiratory chain decreased ATP and increased mitochondrial and cytosolic ROS. Knockdown of CoQ10 biosynthetic genes and complex III genes also increased mitochondrial ROS. In contrast, knockdown of mitochondrial ribosomal proteins (orange) strongly decreased ATP but did not impact ROS, while knockdown of complex IV–associated genes (blue) decreased ATP and mitochondrial ROS. (C) Fast, preranked geneset enrichment analysis reveals pathways of genes that regulate either mitochondrial or cytosolic ROS. Besides the respiratory chain–associated pathways, this analysis highlights ether lipid biosynthesis, which occurs in the peroxisome, as a significant regulator of mitochondrial ROS. Data compiled from n = 2 experiments.

Fig. 3.

Supercomplex and ATP levels are associated with complex I enzyme activity. (A) Cells with complex I subunit knockdown, part of an expanded CRISPRi knockdown library, exhibit a range of mitochondrial ROS or cytosolic ROS phenotypes, as well as a range of ATP phenotypes (previously measured and published in ref. 13), indicated by the color bar. (B and C) Complex I subunits grouped by module. The trendline shows linear regression across complex I subunits and the dotted lines 95% CIs. Subunits are colored based on their association with functional modules within complex I. The notated complex I subunits have prominent effects on ATP and ROS and are examined in follow-up studies. Data are means ± SEM. (D) A representative blue-native PAGE gel, loaded with isolated mitochondria for each of seven cell lines expressing CRISPRi knockdown of complex I subunits and a non-targeting control and stained with total OXPHOS human western blot antibody cocktail. (E) Quantification of n = 3 blue-native PAGE gels shows that knockdown of NDUFB4, NDUFS8, and NDUFA8 decreases supercomplex levels. (F) Quantification of complex I enzyme activity from isolated mitochondria reveals significantly decreased enzyme activity with NDUFA1, NDUFB7, NDUFAB1, NDUFB4, NDUFS8, and NDUFA8 knockdown, from n = 3 replicates. (G) Pearson r correlation matrix of mitochondrial ROS, cytosolic ROS, and ATP phenotypes compiled from 33 complex I subunit knockdowns included in CRISPRi screens from n = 2 replicates, along with complex I activity, supercomplex level, and subcomplex level compiled from seven complex I subunit knockdowns from n = 3 replicates. ATP levels were negatively correlated with mitochondrial ROS and cytosolic ROS and positively correlated with complex I activity. Complex I activity was positively correlated with supercomplex levels. Mitochondrial ROS and cytosolic ROS were also positively correlated. *P < 0.05, **P < 0.01, and ***P < 0.001 by one-way ANOVA with Dunnett’s multiple comparisons test (E and F) and Pearson correlation test (G).

Fig. 4.

Ether lipids bidirectionally control mitochondrial ROS. (A) Peroxisomal and ether lipid synthesis genes and DHODH, whose inhibition elevates peroxisomal ether lipid levels (44), span a wide range of mitochondrial ROS phenotypes, without impacting ATP. Data compiled from n = 2 replicates. (B and C) Representative plot with peak averages and quantification of cells incubated with DHODH inhibitors vidofludimus (Vido, 10 µM), brequinar (Breq, 0.5 µM), or rotenone (Rote, 2 nM) exhibit elevated mitochondrial ROS, measured with MitoSOX. n = 2 replicates with ≥10,000 cells per condition. (D and E) Similar increases in mitochondrial ROS were observed with DHODH inhibitors, measured with MitoNeoD. n = 2 replicates with ≥10,000 cells per condition. (F and G) Representative plot with peak averages and quantification of cells expressing a FRET-based sensor for ATP measured using FACS. There was no consistent difference in ATP in response to 48 h incubation with DHODH inhibitor drugs. ATP-depleted (ATP dep) control cells were treated with oligomycin (10 µM) and 2DG (10 mM). n = 2 replicates with ≥10,000 cells per condition. (H and I) Cells expressing a non-targeting CRISPRi sgRNA (NTG) exhibit elevated mitochondrial ROS with 24 h incubation of 20 µM ether lipid precursor sn-1-O-hexadecylglycerol (OHG). In contrast, K562 cells with CRISPRi knockdown of NDUFS8 (S8) have a diminished mitochondrial ROS response to ether lipid precursor. n = 2 experiments with ≥10,000 cells per condition. ns = not significant, *P < 0.05, and ***P < 0.001 by one-way ANOVA.