Cardiolipin promotes electron transport between ubiquinone and complex I to rescue PINK1 deficiency

Recessive mutations in phosphatase and tensin homologue–induced kinase (PINK1) cause early-onset Parkinson’s disease (PD), an incurable neurodegenerative movement disorder (Valente et al., 2004). When mitochondria are stressed, PINK1 recruits Parkin to promote the autophagic clearance of mitochondria, and genetically PINK1 operates in this context upstream of Parkin (Clark et al., 2006; Park et al., 2006; Narendra et al., 2010). Also under nonstressed, endogenous conditions, the loss of PINK1 results in mitochondrial defects. These defects include reduced ATP levels, cristae defects, and a less negative mitochondrial membrane potential. Important aspects of these phenotypes can be explained by the inability of PINK1 mutants to maintain phosphorylation of serine 250 in subunit NdufA10 of complex I in the electron transport chain (ETC). This defect results in inefficient electron transfer between complex I and ubiquinone in PINK1 mutant fruit fly–, mouse-, and patient-derived cells (Clark et al., 2006; Park et al., 2006; Yang et al., 2006; Gautier et al., 2008; Morais et al., 2009, 2014; Vos et al., 2012). Hence, activation of pathways that increase ETC activity may bypass this electron transfer defect in PINK1 mutants (Vos et al., 2012, 2013) and may constitute a promising therapy for PD.

Changes in lipid content are associated with aging and with neurodegeneration, including Alzheimer’s disease and neurodegeneration with brain iron accumulation (Filippov et al., 2012; Aoun and Tiranti, 2015; Jazvinšćak Jembrek et al., 2015). Likewise, alterations in lipid metabolism have been described in models of PD as well (Lim et al., 2012; Tyurina et al., 2015). However, whether the mitochondrial dysfunction in PINK1 mutants is under the control of specific changes in lipid content has not been investigated, nor has it been thoroughly explored whether mitochondrial dysfunction can be modified by alterations in lipid content.

In this study, we present an unbiased genetic screen and identify fatty acid synthase (FASN) as a dominant suppressor of PINK1 deficiency. Given the strong interest in FASN inhibition in the field of oncology, pharmacological inhibitors exist, and we show that they can also rescue PINK1 deficiency in fly-, mouse-, and patient-derived neurons, connecting the function of FASN to the mitochondrial defects in PINK1 mutants. Interestingly, among several changes in phospholipid levels in PINK1 mutants, we show that the production of the mitochondrial-specific lipid, cardiolipin (CL), is inhibited. Our data indicate that CL, but not other phospholipids, plays a role in the regulation of electron transfer between complex I and ubiquinone. Our work suggests that CL regulates electron transfer in complex I and also indicates that strategies to increase CL levels are protective against mitochondrial dysfunction in PINK1 mutants.

In a genetic screen in fruit flies (Vos et al., 2012), we identified 10 ethyl methanesulfonate–induced mutations that dominantly suppress pink1 mutant phenotypes. When heterozygous, these suppressors rescue decreased ATP levels, mitochondrial morphological defects, and the flying deficit of the pink1^B9-null mutant flies (Fig. 1 A). We used classical deficiency mapping and whole genome sequencing and found a nonsynonymous lesion in 7 of the 10 suppressor mutants (Fig. S1, A and B). One of them, opa1, had been previously identified as a pink1 suppressor (Poole et al., 2008). The strongest pink1 suppressor is a loss-of-function lesion in FASN2 (FASN2/CG3524/v(2)k05816; Fig. 1 A) and is specific to pink1 because heterozygous FASN2 loss only weakly suppresses parkin-null mutant phenotypes (Fig. S1 C). In addition, all the other FASN2 alleles that we obtained also suppress the flight defects and the mitochondrial morphological defects in pink1^B9 mutant flies (Fig. 1, B–F; and Fig. S1, E and F). Note that heterozygous FASN2 animals in an otherwise wild-type background (pink1^RV) did not show defects (Fig. 1, B–F; and Fig. S1, E and F). In addition, expression of wild-type FASN2 rescued the lethality of the different homozygous FASN2 alleles we isolated, further indicating they are loss-of-function mutations (Fig. S1 D). Hence, partial loss of FASN2 rescues pink1 mutant defects.

FASN is an evolutionarily conserved enzyme that catalyzes the synthesis of the free fatty acid palmitate (Fig. 2 A; Chirala and Wakil, 2004). Given the immense interest to inhibit FASN in oncology and metabolic disease (Menendez and Lupu, 2007; Ogino et al., 2008; Nguyen et al., 2010; Park et al., 2015), several inhibitors of this enzyme have been developed (Menendez and Lupu, 2007). Because inhibitors often have several off-target effects, we have tested three independent drugs, Cerulenin, Irgasan, and Orlistat, that block the second, sixth, and seventh steps of the FASN reaction, respectively (Fig. 2 A; Omura, 1976; Kapoor et al., 2004). When 1-d-old adult pink1 mutant Drosophila melanogaster that display flight and mitochondrial defects were placed on medium with one of these compounds, the flies showed a progressive and dose-dependent improvement in flying, and their decreased ATP levels were significantly restored. No significant effect was observed in wild-type control flies that were fed with Cerulenin, Irgasan, or Orlistat (Fig. 2, C, E, and G). Hence, pharmacological inhibition of FASN rescues phenotypes in Pink1-deficient flies, and, in the remainder of our study, we used genetics and Cerulenin.

In addition to a rescue in flight and ATP levels, the characteristic mitochondrial clustering and blebbing defects of pink1 mutant mitochondria in muscles of larvae were rescued after Cerulenin incubation (Fig. 3, A and B). Furthermore, Cerulenin also acutely restores a normal dense mitochondrial cristae structure in pink1 mutant thorax muscles. We placed 1-d-old adult Pink1 mutant animals that show mitochondrial cristae morphology defects on Cerulenin for 3 d and observed that the muscle tissue and mitochondrial morphology and cristae structure of Pink1 mutants that were fed Cerulenin appeared much more like those in control thoraces (with or without Cerulenin), and they show much fewer signs of degeneration (Figs. 3 C and S2, A–C). Interestingly, the rescue we observe is specific for Pink1-induced phenotypes and not for Parkin because 1-d-old Parkin mutant flies that are on Cerulenin for 3 d are not rescued (Fig. S2 D). Although Pink1 and Parkin act together for some aspects of their function (Clark et al., 2006; Park et al., 2006; Narendra et al., 2009; Haddad et al., 2013), the effects of FASN inhibition appear to be specific for pathways under the control of Pink1 (Vilain et al., 2012; Pogson et al., 2014).

To assess whether the rescue of Pink1-associated phenotypes by inhibition of FASN is evolutionarily conserved, we turned to Pink1^−/−-deficient mouse- and patient-derived cells. Incubation of Pink1^−/− mouse embryonic fibroblasts (MEFs) or patient-derived fibroblasts with Cerulenin rescues the decreased ATP levels (Fig. 3, D and E). Compared with age-matched controls, dopaminergic neurons derived from induced pluripotent stem cells (iPSCs) from different PINK1 carriers also displayed significantly lower ATP levels, although this effect was less pronounced than in the fibroblasts. Nonetheless, these lower ATP levels in PINK1 mutant human dopaminergic neurons were also rescued when incubated with Cerulenin (Fig. 3 F). Hence, inhibition of FASN rescues PINK1 deficiency in flies and in mouse- and patient-derived cells.

We used an enzyme-linked assay and measurement of optical density to assess free fatty acid levels that also include free palmitate. Genetic or pharmacological inhibition of FASN in flies and in MEFs showed a trend toward lower free fatty acid levels compared with controls, but the differences were statistically not significant (Fig. 4, A–C, two left bars). We then resorted to gas chromatography–coupled mass spectrometry to directly assess free palmitate levels and found that these were significantly lower in animals treated with Cerulenin (Fig. 4 D, two left bars), in line with the expected result when inhibiting FASN. Surprisingly, pink1 mutants in the absence of FASN inhibition (Fig. 4 D, third bar) also showed a significant reduction in free palmitate levels. These levels were further reduced when Pink1 mutants were treated with Cerulenin (Fig. 4 D, last bar).

The lower levels of free palmitate (and the trend toward lower free fatty acids) in Pink1 mutants may at least partially be explained by our observation that Pink1 mutant cells from mice and patients harbor lower FASN protein levels, as revealed by Western blotting (Fig. 4, E–H). It is well known that Pink1 mutant cells harbor increased reactive oxygen species (ROS; He et al., 2008; Esposito et al., 2013). We therefore tested whether lower cellular FASN levels can be caused by increased ROS, and we treated MEF cells with H₂O₂. This manipulation indeed resulted in lower FASN protein levels (Fig. 4, I and J), indicating that increased ROS correlate with lower FASN protein levels (and thus with lower levels of free palmitate). Although Pink1 mutants already show lower FASN levels and lower free palmitate levels than controls, this reduction is clearly insufficient to rescue the Pink1-associated defects. Yet, as we show here, further lowering FASN (either genetically or pharmacologically, resulting in both cases in lower free palmitate) does yield rescue of the phenotypes.

Palmitate is involved in several downstream pathways, but this lipid also connects to mitochondrial function. Increasing the palmitate levels was shown to inhibit the production of CL, a lipid that is exclusive to the inner mitochondrial membrane and to the membranes of bacteria (Ostrander et al., 2001; Buratta et al., 2008). Unsaturated phosphatidyl glycerol (PG; not saturated PG) is a good substrate for CL synthase (CLS; Hoch, 1992; Rytömaa and Kinnunen, 1994; Bobyleva et al., 1997), and high levels of palmitate shift the balance toward more saturated PG (Fig. 5 A; Ostrander et al., 2001). We used shotgun lipidomics and assessed PG levels in mitochondrial fractions of MEFs. Saturated PG levels were extremely low in controls and were below the detection limit in Pink1 mutant samples. Assessing unsaturated PG levels, we found that inhibition of FASN resulted in a significant increase in most unsaturated PG species (Fig. 5 B). When inhibiting FASN, other lipid species were also expected to be affected. As indicated in Fig. S3, the differences we measured in isolated mitochondria from control cells with and without Cerulenin were not significantly different. Nonetheless, the lower palmitate levels and increased unsaturated PG that we detected upon inhibition of FASN are very robust and consistent with our model.

Pink1 mutants are known to harbor increased mitochondrial ROS (He et al., 2008; Esposito et al., 2013), and high ROS levels are known to result in decreased levels of CL (Ruggiero et al., 1992; Ellis et al., 2005). In line with this, CL levels in mitochondrial fractions of Pink1^−/− MEFs were significantly decreased compared with control MEFs (Fig. 5, C and D). Next, we tested the effect of inhibiting FASN using Cerulenin that causes increased levels of unsaturated PG (see previous paragraph). We found that in Pink1^−/− MEFs, the levels of all CL species measured were almost completely restored to control levels (Fig. 5, C and D). Hence, loss of FASN function results in more mitochondrial CL in Pink1^−/− mutant MEFs.

PINK1^−/− mutants show lower enzymatic activity of complex I (Gautier et al., 2008; Morais et al., 2009). We therefore wondered whether the rescue by inhibiting FASN acts through bypassing this complex I defect. We detected that partial inhibition of FASN using genetic or pharmacological tools restored the defective complex I enzymatic activity in PINK1 mutant flies and in MEFs back to control levels (Fig. 6, A–C). The inefficient electron transport between complex I and ubiquinone in PINK1 mutants was previously shown to be the result of an inability to maintain phosphorylation of Serine 250 in the complex I subunit NdufA10 (Morais et al., 2014). Indeed, mitochondria from engineered mammalian cells expressing only phosphorylation-deficient NdufA10^S250A transfer electrons less efficiently from complex I to ubiquinone, similar to Pink1 mutant mitochondria (Fig. 6 D). Interestingly, we found that pharmacological inhibition of FASN resulted in a restoration of this activity (Fig. 6 D), indicating that the inhibition of FASN bypasses this ETC defect.

Our data are consistent with a model wherein Pink1 mutant mitochondrial dysfunction is rescued because the loss of FASN results in increased mitochondrial CL. We challenged this model by taking three independent approaches. In our first approach, we fed flies an unrelated lipid, stearic acid, that was previously shown to rescue the climbing defect and lower ATP levels seen in Pink1-deficient flies (Senyilmaz et al., 2015). Unlike Cerulenin, stearic acid did not increase the activity of mitochondrial complex I in Pink1 mutants (Fig. S4). These data indicate the specificity of our approach and suggest that stearic acid rescues aspects of the Pink1 phenotype by acting through a different pathway than complex I (e.g., mitochondrial fission; Vilain et al., 2012; Senyilmaz et al., 2015).

In our second approach, we isolated mitochondria from Pink1 mutant flies or Pink1^−/− MEFs and incubated them with different amounts of various lipids, including PG, phosphatidyl ethanolamine (PE), phosphatidyl choline (PC), and CL, while sonicating, thus allowing the lipid to incorporate (Fig. 7, A–D). Only CL, and not PG, PE, or PC, showed a strong rescue of the electron transfer activity defects in complex I of Pink1 mutants (Fig. 7, A–D). This effect was absent in control samples that underwent the same treatment and was also seen when using mitochondrial fractions from Pink1^−/− MEFs (Fig. 7 E). Hence, directly supplementing Pink1 mutant mitochondria with CL can bypass the complex I defect.

In a final approach, we scrutinized our model using genetics. CL is produced in mitochondria by CLS (Nowicki et al., 2005; Acehan et al., 2011), and we observed that partial loss of CLS (CLS^e01021/+) in fruit flies resulted in lower complex I activity compared with the complex I activity measured in control flies (Fig. 7 F). This defect in complex I activity in CLS/+ animals was not rescued by the inhibition of FASN with Cerulenin (Fig. 7 F, first green bar). In addition, the rescue of complex I defects in Pink1 mutants by Cerulenin is lost in Pink1 mutants that are also mutants for CLS (Fig. 7 F, right green bar). These data further support the model wherein the rescue of complex I defects in Pink1 mutants by the inhibition of FASN follows a CL-dependent pathway.

CL stabilizes cristae and ETC supercomplexes (Pfeiffer et al., 2003; Friedman et al., 2015), but a role for this lipid in the direct regulation of electron transport between complex I and ubiquinone in animal mitochondria has not been assessed. As noted in the Loss of FASN increases complex I function section, electron transfer in Pink1 mutant mitochondria is caused by defective NDUFA10 phosphorylation (Morais et al., 2014). We therefore tested the ability of CL to bypass the electron transfer defect in NdufA10^S205A phosphorylation-deficient cells. We incubated mitochondria from such cells with CL and found that this manipulation efficiently rescued the complex I defects (Fig. 7 G). Hence, CL facilitates electron transport between complex I and ubiquinone, and increasing CL levels are sufficient to bypass the defects in Pink1 and NdufA10^S205A mutants.

In this study, we present a previously uncharacterized function for the mitochondrial inner membrane lipid CL, and we show that increased CL rescues Pink1-induced mitochondrial dysfunction. CL has been coupled to the maintenance of mitochondrial cristae structure, and it is key to maintaining ETC supercomplexes (Pfeiffer et al., 2003). We now expand on this and provide evidence that CL also couples electron transfer between complex I and ubiquinone to increase the efficiency of the ETC. Pink1 mutants and many sporadic PD patients suffer from such complex I defects. We find that CLS mutants show decreased electron transfer from complex I to ubiquinone; conversely, increasing CL levels by partial inhibition of FASN using genetics or pharmacology significantly rescues the mitochondrial defects in Pink1 mutant cells (in both mice and patients) as well as in the degenerative phenotypes and flight defects in adult fruit flies. Loss of FASN also affects the levels of other lipids in isolated mitochondria. However, we did not observe an improvement in complex I activity after the incubation of isolated mitochondria with these lipids, suggesting that these changes in lipid level likely are not significantly contributing to the PINK1 mutant phenotypes, as we report in this study.

Thus, our work also indicates specificity to CL, as the biochemical defect at complex I in Pink1 mutants is not affected by other lipids, including PG, PE, PC, and stearic acid.

Our data are consistent with a significant contribution of ETC defects to Pink1 pathology, but they also confirm that Pink1 acts in multiple parallel pathways. Stearic acid was recently shown to rescue Pink1-associated defects in fruit flies by activating pathways that induce mitochondrial fission (Senyilmaz et al., 2015). We and others previously showed that mitochondrial fission acts in parallel to a role of Pink1 in regulating complex I activity because increasing mitochondrial fission in Pink1 mutants does not rescue their complex I defect (Vilain et al., 2012; Pogson et al., 2014). Similar to these findings, stearic acid treatment does not rescue the complex I defect of Pink1 mutants. These observations are in contrast to those of increasing CL levels: CL does rescue the complex I dysfunction, but our EM study shows that mitochondria still appear somewhat enlarged, as was also seen previously when expressing a phosphomimetic NDUFA10^S250D in Pink1 mutant animals (Morais et al., 2014). Although targeting either pathway significantly improves mitochondrial dysfunction such that the mutant flies can fly again, a strategy that targets both simultaneously may yield even stronger effects (Vilain et al., 2012).

Neurodegeneration and aging have been linked to changes in lipid composition. The lower CL levels may be the consequence of mitochondrial stress, including increased ROS that accumulate because of inefficient electron transport in complex I (He et al., 2008; Esposito et al., 2013). Although it is recognized that upon mitochondrial stress, CL is relocated to the outer mitochondrial membrane (Garcia Fernandez et al., 2002) to serve as a recognition signal for the degradation of mitochondria (Chu et al., 2013), the mitochondrial mass in Pink1 mutant flies or cells is not particularly lower compared with wild-type controls (Rakovic et al., 2013). Rather, the defective mitochondria in Pink1 mutants that are not degraded can be rescued by increasing CL levels, and this manipulation appears sufficient and specific to restore a dense cristae structure and to bypass the complex I dysfunction in Pink1 mutants. Other phospholipids were not able to rescue the complex I defects like CL could, suggesting that these phospholipids do not significantly contribute to the rescue of complex I dysfunction in PINK1 deficiency.

The loss of Pink1 also correlates with lower palmitate levels. The production of palmitate by FASN requires a considerable amount of energy (Wakil et al., 1983), a feat particularly relevant in the context of Pink1 mutants that harbor lower cellular ATP levels (Clark et al., 2006; Park et al., 2006). Pink1-defective cells may thus access other sources in an attempt to maintain free fatty acid levels. The lower FASN levels we observed in Pink1 mutants and in H₂O₂-treated cells may be a protective mechanism under such conditions of low cellular energy, but further work is needed here. The palmitate mass that we measured in Pink1 mutant samples may thus not be a simple reflection of direct cellular palmitate synthesis by FASN but a combination of different contributing factors, such as lower uptake, less synthesis, and more consumption. These findings support the idea that inhibiting FASN in Pink1 mutants not only restricts palmitate production, thus allowing more CL synthesis, but that this maneuver may, to some extent, also help to conserve cellular energy.

Thus, we propose that the lower FASN levels or activity may help to conserve energy and, more significantly, result in increased levels of CL to facilitate electron transfer between complex I and ubiquinone. FASN is a major target in oncology and metabolic disease, and phase I trials to inhibit the enzyme are ongoing. However, unlike the need to strongly inhibit the enzyme in oncology applications, our genetic work predicts that in PD, partial inhibition is sufficient, thereby reducing the risk of side effects.

w pink1^B9 (FBgn0029891) and park¹ (FBgn0041100) mutant and w pink1^RV and park^RV control flies were provided by J. Park and J. Chung (Korea Advanced Institute of Science and Technology, Daejeon, South Korea; Cha et al., 2005; Park et al., 2006), park^Δ21 mutants were provided by G. Mardon (Baylor College of Medicine, Houston, TX; Pesah et al., 2004), and UAS-Fasn2 was provided by J. Montagne (Université Paris-Sud, Orsay, France; Garrido et al., 2015). Deficiencies and P-element mutants were from the Bloomington Drosophila Stock Center (Bier et al., 1989; Ashburner et al., 1990; Spradling et al., 1999; Mason et al., 2004; Parks et al., 2004; Ryder et al., 2007). dFasn2¹ is y w eyFLP.GMR::LacZ; dFasn¹ P[y⁺] P[FRT40A neo⁺] and control is y w eyFLP GMR::LacZ; P[y⁺] P[FRT40A neo⁺] (Verstreken et al., 2005). k05816 (FBgn0042627) is y¹ w^67c23; P{lacW}FASN2^k05816/CyO and EP695 is w¹¹¹⁸; P{EP}FASN2^EP695/CyO. The 10 ethyl methanesulfonate screen mutants are y w eyFLP GMR::LacZ; P[y⁺] P[FRT40A neo⁺] *mutant (Verstreken et al., 2005). CLS flies are w¹¹¹⁸; PBac{RB}CLS^e01021/TM6B, Tb¹.

Pink1^−/− MEFs were previously described (Morais et al., 2009). In brief, heterozygous Pink1 knockout mice (Pink1^+/−) were interbred to generate mutant mice and wild-type littermate controls. At embryonic day 13, embryos were dissected into cell culture dishes containing DMEM supplied with 50% FCS and 1% penicillin/streptomycin. Cultures were expanded and serum concentration was constantly decreased to 10% FCS after they reached consistent growth. Afterward, the cultures were immortalized by transfection with SV40 large T antigen.

The NdufA10 HeLa cells were described and characterized previously (Morais et al., 2014). In brief, the HeLa NdufA10 down-regulated cell line was generated using shRNA against human NdufA10 (targeted sequence, 5′-CAGAAGAAAGGAGATCCACATGAAATGAA-3′) according to the supplier’s protocol (OriGene) and were selected based on their acquired puromycin resistance.

We have studied fibroblast cultures from two PINK1 mutation carriers harboring homozygous p.Q456X and p.V170G mutations (female, mean age 80 ± 0.7 yr; mean age at onset: 46 ± 21 yr). In addition, two cultures from healthy control individuals were included (female, mean age 51 ± 4.9 yr) that are family members from the patient carrying the p.Q456X mutation (Grünewald et al., 2009).

The iPSC lines were differentiated into dopaminergic neurons as described previously (Kriks et al., 2011; Munsie et al., 2015). In brief, iPSC colonies were dissociated into single cells using accutase (Thermo Fisher Scientific) and replated onto matrigel-coated dishes (BD). Once iPSCs reached a confluence of 100%, neural differentiation was initiated by adding knockout serum replacement medium supplemented with the SMAD pathway inhibitors SB431542 (Tocris Bioscience) and LDN-193189 (Stemgent). During the first 20 d, several growth factors and small molecules were added to the cell culture medium at various time points as indicated in the previously published protocol (Kriks et al., 2011) to pattern the neurons into the dopaminergic lineage (floor plate induction). On day 21 of the differentiation, cells were dissociated using accutase and plated at a high density (50-µl drop with 350,000 cells) onto the center of culture dishes, which had been previously coated with poly-d-lysine (Sigma-Aldrich) and laminin (Roche), in medium for terminal differentiation (Kriks et al., 2011). Neurons were aged for ∼50 d.

1-d-old adult flies were placed on molasses medium supplemented with 100 µM Cerulenin (Enzo Life Sciences), 3 µM Irgasan (Sigma-Aldrich), or 100 µM Orlistat (Sigma-Aldrich), unless otherwise indicated in Fig. 2 (B, D, and F; 2.5% final ethanol concentration), or flies were supplemented with 2.5% ethanol for control medium. The animals were kept on these media for 72 h and tested for flight or ATP levels. To determine mitochondrial morphology, embryos were placed in molasses medium supplemented with 100 µM Cerulenin (2.5% final ethanol concentration) or with 2.5% ethanol for control medium and grown to the third instar larval stage. These third instar larvae were dissected and used for experiments (see Determination of mitochondrial morphology).

To assess ATP levels in the cells, cells were treated with 10 µM Cerulenin for 24 h (0.1% ethanol) or with control medium supplemented with 0.1% ethanol. For Western blotting analysis, wild-type MEFs were treated with 100 µM H₂O₂ for 30 min.

The flight assay was conducted on male flies using batches of five flies each. The flies were placed in an empty vial and gently tapped. The flies that were able to fly were given a score of 1, whereas those that did not fly were given a score of 0.

For survival, the latest survival stage of animals is indicated for animals that were reared on grape juice plates with yeast paste in uncrowded conditions.

ATP levels were determined as previously described (Park et al., 2006; Morais et al., 2014) using an ATP determination kit (Invitrogen) according to the supplier’s protocol. Luminescence was measured on a luminometer (Bio-Rad Laboratories), and values were normalized to total protein content using the Bradford method.

Flies and cells were homogenized in 1-ml chilled mitochondrial isolation medium (Abcam) and then spun twice at 1,000 g for 5 min. The supernatant was collected and spun at 13,000 g for 5 min followed by a washing step of the pellet containing the mitochondria with 1 ml of mitochondrial isolation medium and was resuspended in 50 µl isolation medium. The samples were kept at 4°C during the entire process (Walker et al., 2006; Frezza et al., 2007). Mitochondrial isolates were sonicated (10 pulses at 20%) and used to measure complex I activity as previously described (Fischer et al., 1986; de Paepe et al., 2006). In brief, to measure NADH:O₂ oxidoreductase activity, mitochondrial isolates were buffered in medium containing phosphate buffer (0.1 M, pH 7.4), BSA (10 mg/ml), NADH (1 mM), and KCN (30 mM). Decylubiquinone, a ubiquinone analogue, was added to accept the electrons from complex I and to activate the reaction. The activity of NADH oxidation was spectrophotometrically followed at 340 nm at 37°C and was corrected for rotenone-insensitive NADH oxidation. Values were normalized to citrate synthase activity.

For phospholipid incubation, isolated mitochondria underwent freeze-thaw steps and sonication and were then incubated for 60 min with various concentrations of the different lipids at 37°C before complex I activity measurements. The phospholipid analysis revealed that the monounsaturated fatty acyl chains are most abundant (except for CL). Therefore, the following phospholipids were used: CL (LM-6003), PC (850457C), PG (840457C), and PE (850757C).

Cells were collected and diluted in Laemmli sample buffer and were boiled for 5 min. 15 µg of proteins were separated on precast 4–12% NuPage Bis-Tris gels (Invitrogen). Blots were probed with the following primary antibodies: 1:1,000 rabbit anti-FASN and 1:1,000 anti–β-actin (Abcam). The membranes were developed with Western Lightning ECL (PerkinElmer) and imaged. Quantification was performed using the gel analyzer tool in ImageJ software (National Institutes of Health).

Larvae were dissected in HL-3 (Stewart et al., 1994) and larval fillets were fixed in 4% formaldehyde in PBS for 20 min and permeabilized with 0.4% Triton X-100. Primary antibody monoclonal mouse anti-ATP synthase B (1:200; Abcam) was used to visualize mitochondria. We used goat anti-mouse Alexa Fluor 555 (1:1,000) as a secondary antibody. Images were visualized with a confocal microscope (A1R) through a near-infrared Apo 60× 1.0 NA water dipping lens (Nikon) in room temperature. Images were acquired using the NIS-Elements Advanced Research software package (Nikon). Images were quantified using the analyzing particles plugin in ImageJ, with which rounded mitochondria were automatically detected and counted.

Electron microscopic analyses of thorax mitochondria were performed as previously described (Esposito et al., 2013). In brief, thoraxes of adult flies were fixed in paraformaldehyde/glutaraldehyde, postfixed in osmium tetroxide, dehydrated, and embedded in epon. 80-nm-thick sections were stained with uranylacetate end lead citrate, sections were imaged using a transmission electron microscope (TEM 1400; JEOL) at room temperature and a bottom-mounted camera (11 MP; Quemesa; Olympus), and iTEM 5.2 software (ResAlta Research Technologies) was used to acquire micrographs. Mitochondrial morphology was analyzed using ImageJ and scored in a blinded fashion as normal mitochondria with organized cristae or as swollen mitochondria devoid of cristae from five flies. For each fly, five sections were analyzed.

30 adult flies were homogenized in 5% isopropanol and 5% Triton X-100 and prepared according to the manufacturer’s protocol (BioAssay Systems). The amount of free fatty acids was analyzed by measuring optic density at 570 nm on a Multilabel reader (Biotek). Values were normalized to protein content.

One million cells were plated in 6-well plates and incubated with Cerulenin for 24 h before sample collection. Metabolites for the subsequent mass spectrometry analysis were prepared by quenching the cells in liquid nitrogen followed by a cold two-phase methanol/water/chloroform extraction (Christen et al., 2016). During the extraction, samples were constantly refrigerated in a dry ice/ice mixture. Each sample was scraped in 800 µl of a 3:5 water/methanol solution and collected in a 1.5-ml tube. Norvaline (0.75 µg/sample) and glutarate (0.75 µg/sample) were added as internal standards for polar metabolites. Next, 500 µl chloroform was added to each sample. Heptadecanoate (5 µg/sample) was also added as an internal standard for fatty acids. Samples were vortexed for 10 min at 4°C, and phase separation was then achieved by centrifugation at 4°C at maximum speed for 10 min on a refrigerated bench-top centrifuge. The water/methanol phase containing polar metabolites, the chloroform phase containing fatty acids, and the interphase containing proteins and DNA were separated and dried using a vacuum concentrator. Dried metabolite samples were stored at −80°C until further processing. Fatty acids were esterified with 2% sulfuric acid in methanol, shaken for 15 min at room temperature, and subsequently extracted by addition of 600 µl hexane and 100 µl of saturated NaCl per sample, followed by vortexing for 10 min and centrifugation at 4°C at maximum speed for 5 min on a refrigerated bench-top centrifuge. The hexane (upper) phase was transferred to new 1.5-ml tubes and dried using a vacuum concentrator. Samples were then resuspended in 25–100 µl hexane, vortexed for 10 min, and transferred to glass vials for gas chromatography–mass spectrometry analysis. The method was tested on free fatty acid and triglyceride standards and was verified to specifically esterify free fatty acids. Metabolite levels were measured with a gas chromatography system (7890A) combined with an inert mass spectrometry system (5975C; Agilent Technologies) as described previously (Lorendeau et al., 2016). Specifically, 1 µl of sample was injected into a DB35MS column (Agilent Technologies) in splitless mode using an inlet temperature of 270°C. The carrier gas was helium with a flow rate of 1 ml/min. Upon injection, the gas chromatography oven was held at 80°C for 1 min and then ramped with 5°C/min to 300°C. The mass spectrometry system was operated under electron impact ionization at 70 eV, and a mass range of 100–650 atomic mass units was scanned. The palmitate peak was identified based on a standard curve, and the area under the curve of the palmitate peak was integrated using a MatLab M file, published by Young et al. (2008), which applies consistent integration bounds and baseline correction to each ion. The total ion counts for palmitate of each sample was compared with a blank extraction without cells to correct for the eventual presence of noncellular tracers of palmitate. The data were normalized over total DNA content. For this purpose, total DNA was extracted from the interphase with the DNeasy Blood and Tissue kit (QIAGEN) and quantified with a spectrophotometer (Nanodrop 2000; Thermo Fisher Scientific).

Mass spectrometry–based lipid analysis was performed at Lipotype GmbH as described previously (Sampaio et al., 2011). Lipids were extracted using a two-step chloroform/methanol procedure (Ejsing et al., 2009). Samples were spiked with internal lipid standard mixture containing CL 16:1/15:0/15:0/15:0, ceramide 18:1:2/17:0, diacylglycerol 17:0/17:0, hexosyl ceramide 18:1:2/12:0, lysophosphatidate 17:0, lysophosphatidyl choline 12:0, lysophosphatidyl ethanolamine 17:1, lysophosphatidyl glycerol 17:1, lysophosphatidyl inositol 17:1, lysophosphatidyl serine 17:1, phosphatidate 17:0/17:0, PC 17:0/17:0, PE 17:0/17:0, PG 17:0/17:0, phosphatidyl inositol 16:0/16:0, phosphatidyl serine 17:0/17:0, cholesterol ester 20:0, sphingomyelin 18:1:2/12:0:0, triacylglycerol 17:0/17:0/17:0, and cholesterol D6. After extraction, the organic phase was transferred to an infusion plate and dried in a speed vacuum concentrator. First-step dry extract was resuspended in 7.5 mM ammonium acetate in chloroform/methanol/propanol (1:2:4; V:V:V), and second-step dry extract was resuspended in 33% ethanol solution of methylamine in chloroform/methanol (0.003:5:1; V:V:V). All liquid handling steps were performed using the STARlet robotic platform with the Anti-Droplet Control feature for organic solvents pipetting (Hamilton Robotics).

Samples were analyzed by direct infusion on a mass spectrometer (QExactive; Thermo Fisher Scientific) equipped with an ion source (TriVersa NanoMate; Advion Biosciences). Samples were analyzed in both positive and negative ion modes with a resolution of mass/charge ratio = 200 = 280,000 for mass spectrometry and Rm/z = 200 = 17,500 for tandem mass spectrometry experiments in a single acquisition. tandem mass spectrometry was triggered by an inclusion list encompassing corresponding mass spectrometry mass ranges scanned in 1-D increments (Surma et al., 2015). Both mass spectrometry and tandem mass spectrometry data were combined to monitor CE, diacylglycerol, and triacylglycerol ions as ammonium adducts; PC and PC O⁻ as acetate adducts; and CL, phosphatidate, PE, PE O⁻, PG, phosphatidyl inositol, and phosphatidyl serine as deprotonated anions. Mass spectrometry only was used to monitor lysophosphatidate, lysophosphatidyl ethanolamine, lysophosphatidyl ethanolamine O⁻, lysophosphatidyl inositol, and lysophosphatidyl serine as deprotonated anions; ceramide, hexosyl ceramide, sphingomyelin, lysophosphatidyl choline, and lysophosphatidyl choline O⁻ as acetate adduct; and cholesterol as an ammonium adduct of an acetylated derivative (Liebisch et al., 2006).

Data were analyzed with in-house developed lipid identification software based on LipidXplorer (Herzog et al., 2011, 2012). Data postprocessing and normalization were performed using an in-house developed data management system. Only lipid identifications with a signal-to-noise ratio of >5 and a signal intensity fivefold higher than in corresponding blank samples were considered for further data analysis.

Statistical analyses was performed using Prism software (GraphPad), and to assess whether the observed data points were significantly different, we used a two-tailed Student’s t test when comparing two conditions or one-way analysis of variance with Bonferroni’s post-hoc test when comparing more than two conditions. P-values were considered significant when they were <0.05.

Fig. S1 shows the identification of the suppressive alleles in pink1^B9 mutant flies. Fig. S2 shows that inhibition of FASN results in the suppression of pink1^B9 mutant flies, but not in the suppression of Parkin mutant flies. Fig. S3 shows the levels of various phospholipids in Pink1^−/− MEFs upon inhibition of FASN. Fig. S4 shows the complex I activity of pink1 mutant flies that were pretreated with stearic acid.

We thank G. Marschner, F. Rudolph, R. Goodchild, H. Bellen, J. Park, J. Chung, and the Bloomington Drosophila Stock Center.

Support was provided by a Fonds Wetenschappelijk Onderzoek and European Molecular Biology Organization stipend to M. Vos (ALTF 851-2014), a European Research Council Starting and Consolidator grant to P. Verstreken (ERC-2014-CoG–646671), Fonds Wetenschappelijk Onderzoek grants (G0D3417N, G0D3317N, G088515N, and G094915N), an Interuniversitaire Attractie Pool by Federaal Wetenschapsbeleid (P7/16), the Research Fund Katholieke Universiteit Leuven, a Methusalem grant from the Flemish government, the Hercules Foundation, Opening the Future, Deutsche Forschungsgemeinschaft (KL1134/11-1), and the Vlaams Instituut voor Biotechnologie. B. De Strooper is a paid consultant for and P. Verstreken receives research funding from Janssen Pharmaceuticals. C. Klein is the recipient of a career development award from the Hermann and Lilly Schilling Foundation.

The authors declare no further competing financial interests.

Author contributions: Conceptualization, methodology, and writing: M. Vos and P. Verstreken; investigation: M. Vos, A. Geens, C. Böhm, L. Deaulmerie, J. Swerts, M. Rossi, K. Craessaerts, E.P. Leites, P. Seibler, A. Rakovic, T. Lohnau, and V.A. Morais; supervision, review, and editing: M. Vos, B. De Strooper, S.-M. Fendt, V.A. Morais, C. Klein, and P. Verstreken.