The Drosophila inner-membrane protein PMI controls crista biogenesis and mitochondrial diameter

Mitochondria are present in all eukaryotic cells where they play crucial functions including energy supply, apoptosis induction and calcium buffering (Nunnari and Suomalainen, 2012). Although involved in processes common to all cells, mitochondria are strikingly polymorphic organelles whose shape, length and diameter vary between cell types. Mitochondrial architecture is highly influenced by physiological and pathological conditions, highlighting the functional links between mitochondrial morphology and activity (Dürr et al., 2006; Detmer and Chan, 2007; Suen et al., 2008; Mitra et al., 2009; Tondera et al., 2009; Mitra et al., 2012; Nunnari and Suomalainen, 2012). If it is well established that mitochondria morphogenesis is supported by fusion and fission events regulated by dynamin-related GTPases (Mitofusin and DRP1) (Okamoto and Shaw, 2005; Westermann, 2010), the mechanisms by which mitochondria adopt their tubular shape, and regulate their diameter remain largely unknown. Recent evidence indicates that inner-membrane structures and their associated proteins are part of these processes (Dimmer et al., 2005; Okamoto and Shaw, 2005; Rabl et al., 2009; Velours et al., 2009; Zick et al., 2009; Head et al., 2011).

Mitochondria inner membranes are compartmentalized into domains such as inner-membrane boundaries, crista junctions and crista tubules that, by expressing specific proteins, support specific function (Frey and Mannella, 2000; Mannella, 2006a; Zick et al., 2009). Inner-membrane boundaries are contact sites between the inner and outer membranes that are maintained through physical links tethering the inner-membrane TIM complex to its outer-membrane counterpart, TOM (Vogel et al., 2006; Wurm and Jakobs, 2006). Cristae are invaginations that appear as straight and parallel tubular structures evenly distributed along the mitochondrial length (Mannella, 2006b; Zick et al., 2009). The opening of cristae onto the inter-membrane space, referred to as cristae junctions, consist of narrow tubules, which create a cristae-specific molecular environment by preventing the diffusion of membrane and soluble molecules (Frey and Mannella, 2000; Zick et al., 2009). Few molecular determinants of crista morphogenesis have been characterized in details. One of them, Mitofilin, assembles into the MINOS/MitOS/MICOS complex to support the formation of crista junctions (John et al., 2005; Xie et al., 2007; Rabl et al., 2009; Harner et al., 2011; Hoppins et al., 2011; von der Malsburg et al., 2011). OPA1, originally identified as a key effector of inner-membrane fusion, also controls the tightening of crista junctions (Olichon et al., 2003; Cipolat et al., 2006; Frezza et al., 2006; Meeusen et al., 2006). The non-catalytic subunits e and g of the F0F1 ATP-synthase (F0F1 e/g su) support the positive bending of crista tips (Strauss et al., 2008; Rabl et al., 2009). Finally, the regulation of inner-membrane component production also impacts on crista morphogenesis. Consistently, mutations in the cardiolipin synthase tafazzin result in inner-membrane defects (Xu et al., 2006; Acehan et al., 2007), whereas loss of Prohibitin that affects inner-membrane phospholipid content inhibits crista growth (Osman et al., 2009b; Osman et al., 2009a).

Consistent with their role in crista organization, proteins coding for crista components indirectly control mitochondrial oxidative metabolism (John et al., 2005; Rabl et al., 2009), growth (John et al., 2005; von der Malsburg et al., 2011) and DNA content (Paumard et al., 2002), and their dysfunction has an impact at the organism level (Xu et al., 2006; Mun et al., 2010; Head et al., 2011). In human, abnormal inner-membrane organization is a common trait for many neuromuscular and metabolic diseases including Barth's syndrome (Acehan et al., 2007), Leigh's syndrome (Celotto et al., 2006), Alzheimer's disease (Baloyannis, 2006), amyotrophic lateral sclerosis (Chung and Suh, 2002), Wolfs Hirschhorn's syndrome (Hasegawa and van der Bliek, 2007; Dimmer et al., 2008) and polymyocitis (Zick et al., 2009).

In this manuscript, we described the consequences associated with an in vivo inactivation of Drosophila Pantagruelian Mitochondrion I (PMI), an inner-membrane protein that regulates mitochondrial morphogenesis (Rival et al., 2011). We show that loss of PMI function induces crista elongation and disorganization in Drosophila cells. We demonstrate that these inner-membrane abnormalities are associated with decreased oxidative metabolism and increased ROS production. We further show that at the organism scale, PMI mutation provokes synaptic defects and premature death of Drosophila adults. Altogether our results suggest that, by controlling crista biogenesis, PMI determines mitochondrial diameter explaining why PMI mutation also affects the overall tubular shape of mitochondria.

Several reports have highlighted the importance of inner-membrane proteins in mitochondrial crista morphogenesis (Paumard et al., 2002; Rabl et al., 2009; Harner et al., 2011; Hoppins et al., 2011; von der Malsburg et al., 2011). Having shown that PMI is an inner-membrane protein involved in the regulation of mitochondrial shape (Rival et al., 2011), we tested whether its inactivation would impact crista morphology in Drosophila embryos. In contrast to control mitochondria that form connected elongated tubules, mitochondria of PMI mutants form individualised spheres (Fig. 1A). Quantification of mitochondrial shape parameters on electron microscopy sections revealed that PMI mutant mitochondria contain significantly longer cristae than control mitochondria (Fig. 1C,D). This increase in length was correlated with an increase of the crista membrane length over outer membrane length ratio (c.m./o.m.; Fig. 1F). Conversely, we showed that PMI overexpression (da-GAL4 UAS-PMI) triggers a reduction of the average crista length and a decrease of the c.m./o.m. ratio (Fig. 1C,D,F). Despite this increased biogenesis, the overall morphology of the cristae was unaffected by PMI inactivation. As for wild type, PMI mutant cristae were quite straight and parallel (Fig. 1C4,G), although 10% of mitochondria display curved cristae suggesting that curvature is rather a consequence than a cause of increase length (Fig. 1C5). In addition, the diameter of PMI crista tubules was unaffected (Fig. 1H), and crista junctions were still present in PMI mutant mitochondria (Fig. 1I). We then compared PMI mitochondrial phenotype to that of drp1 and opa1 that are, respectively, effectors of mitochondrial fission and inner-membrane fusion. Consistent with the consequence of unbalanced fusion drp1 mutant embryos display interconnected hanks of long tubular mitochondria (Fig. 1A). Conversely, the invalidation of the pro-fusion gene opa1 results in the fragmentation of mitochondria into numerous small spherical entities (Fig. 1A). In contrast to PMI mutants, in drp1 and opa1 mutant mitochondria the crista length and c.m./o.m. ratio was normal (Fig. 1C,D,F). Similar results were obtained in adult brains in which PMI inactivation results in unique giant mitochondria in neurons instead of the thin tubules observed in the wild type (supplementary material Fig. S1A,B; Fig. S2A) (Rival et al., 2011). While mitochondria of wild-type neuron displayed straight, parallel and evenly spaced cristae (Fig. 2A,B), cristae from PMI mutant neurons appeared dramatically bent and longer than those in control mitochondria (Fig. 2A,B). At the mitochondrion periphery, PMI mutant inner-membrane invaginations often form concentric swirls that were never observed in controls (Fig. 2A,B). Although less pronounced, similar elongation and bending of crista membranes were also present in mitochondria from adult flight muscles (Fig. 3). To determine whether bending of PMI cristae was not a simple consequence of the overall increased size of mitochondria, we analysed drp1 mutant mitochondria. As reported for drp1 knockout mouse (Ishihara et al., 2009), drp1 mutant mitochondria from adult fly neurons form giant masses, which are morphologically indistinguishable from those of PMI mitochondria (Rival et al., 2011). Although of same size and shape as PMI mitochondria, drp1 mitochondria still contains straight and parallel cristae that are regularly distributed within the mitochondrial matrix as in wild-type cells (Fig. 1A). Taken together our results demonstrate that the inner-membrane protein PMI controls the elongation and the architecture of crista membranes.

In PMI mutant embryos, the average diameter of mitochondria is strongly increased compared to controls, a phenotype that is not observed in drp1 or opa1 mutants (Fig. 1B). To test whether the uncontrolled crista biogenesis seen in PMI mutants was related to increased mitochondrial diameter, we quantified these parameters in mutant embryos. These quantifications highlighted a direct correlation between crista length and mitochondrial diameter at this early developmental stage (Fig. 1E). Later, at the adult stage, PMI inactivation in adult brain neurons (Fig. 2A; supplementary material Fig. S1B; Fig. S2A) and in flight muscles (Fig. 3; supplementary material Fig. S3) was similarly associated with wider mitochondria. Conversely, PMI overexpression in adult flight myocytes was sufficient to trigger a dramatic reduction of mitochondrial diameter and length (mef2-GAL4 UAS-PMI; Fig. 3; supplementary material Fig. S3). This contrasted with drp1 overexpression or knockdown of opa1, which reduced mitochondrial length without affecting the diameter (Fig. 3; supplementary material Fig. S3). Cristae from PMI-overexpressing myocytes appeared shorter and less packed than in wild-type mitochondria, consistent with a decreased in crista membrane biogenesis (Fig. 3). Again this contrasted with the wild-type cristae observed in drp1-overexpressing cells, or with the vacuolated and highly disorganised cristae from opa1 knockdown or overexpressing myocytes (Fig. 3). Altogether, these results suggest that, by controlling the biogenesis of crista membranes, the inner-membrane protein PMI can influence the diameter and in turn the tubular shape of mitochondria.

As we showed that PMI is required for normal crista organisation, we investigated whether PMI inactivation would affect the oxidative metabolism that takes place in the mitochondrial cristae. As shown in Fig. 4, ATP levels were significantly lower in extracts from PMI mutant than from control flies, a phenotype that could be rescued by genetically restoring PMI expression (Fig. 4A). A similar reduction was observed after treatment of the flies with the respiratory chain blocker Paraquat. We then quantified mitochondrial membrane potential (ΔΨm) using JC1 dye that accumulates in the mitochondrial matrix as green monomers and aggregates into red oligomers proportionally to ΔΨm strength (Smiley et al., 1991; Reers et al., 1995). As expected, treatment with the proton ionophore DNP strongly inhibits the formation of JC1 aggregates in adult flight muscle mitochondria (Fig. 4C). Similarly, loss of PMI function was associated with a reduction of JC1 aggregates in muscle mitochondria, a phenotype that could be rescued by restoring PMI expression (Fig. 4B,C). Since the proton gradient is generated by the redox activity of the electron transport chain (ETC) complexes, we measured state III respiration in adult fly extracts, and found that the activity of mitochondrial complex I and II was significantly decreased in PMI mutants (Fig. 4D). This correlated with a decrease in the protein levels of ETC subunits from complex I and II in the absence of transcriptional repression, loss of mitochondrial DNA, or downregulation of the inner-membrane-associated protein ATP-synthase (supplementary material Fig. S4). Finally, as respiratory chain defects could result in the release of reactive oxygen species (ROS), we compared ROS production in wild-type and PMI mutants using the dihydroethidium (DHE) dye whose fluorescence is directly proportional to the level of superoxide free radicals. ROS levels measured in PMI mutant adult fly brains (Fig. 4E,F) were higher than in controls, consistent with results obtained with brains treated with Paraquat. It is noteworthy that in drp1 mutants mitochondria, which are as large as PMI mitochondria but whose crista organisation is normal, ATP levels, ΔΨm strength and ROS levels were not affected (Fig. 4A,C,F). This suggested that the alterations in ATP levels, ROS production and ΔΨm detected in PMI mutant mitochondria are a consequence of altered crista shape and architecture rather than an effect of the overall mitochondrial size increase. Taken together, these results demonstrate that respiratory chain function relies on PMI-dependent crista morphogenesis.

Having characterised the consequences of the PMI mutation at the subcellular level and more specifically on mitochondrial shape and structure, we analysed the impact of inactivating PMI function at the organism level. The first noticeable phenotype was a reduction of the lifespan of PMI mutants. Whereas the median survival of wild-type animals reaches 50 days, it dropped to 13 days in PMI mutant flies, a phenotype greatly improved by restoring PMI expression (Fig. 5A). This premature death was not due to increased mitochondria size since drp1 inactivation did not result in such premature adult lethality (Fig. 5A). In addition, the fact that brains and muscles from aged PMI flies did not show any sign of degeneration (vacuolization, apoptotic nuclei) or mitochondrial damages (supplementary material Fig. S1A–C), led us to suspect that adult lethality could be secondary to the increased ROS levels. This hypothesis was validated by experiments showing that overexpression of the mitochondrial specific anti-oxidant protein Mn-superoxide dismutase (SOD2) significantly increased median survival of PMI mutant (Fig. 5B). These results demonstrate that PMI mutant premature lethality is partly the consequence of excessive ROS production by PMI mutant mitochondria. It should be noted that SOD2 overexpression did not rescue mitochondrial defects, indicating that abnormal mitochondrial morphology is not caused by an excess of ROS (supplementary material Fig. S5). These results demonstrate that PMI-dependent alterations at the mitochondrial level dramatically impair survival at the organism level.

We then tested whether PMI loss-of-function mutation could impair neuronal activity, which strongly relies on mitochondria for energy supply and calcium buffering. Behavioural and locomotion tests were applied to both wild-type and PMI mutant flies. When performing these tests, we observed that PMI mutants were paralyzed when submitted to a stressful mechanical shock. This phenotype, known as bang sensitive, has been described for Drosophila mutants affecting certain aspect of mitochondria function (Fig. 6A; supplementary material Movie S1). We therefore measured the synaptic activity at the neuromuscular junction of PMI mutant flies using a well-established protocol (Tanouye and Wyman, 1980; Elkins and Ganetzky, 1990; Dupont et al., 2012). Although PMI mutants were able to generate excitatory post-synaptic potential (EPSP), a higher rate of EPSP failure was detected in these flies (Fig. 6B,C). Both bang sensitivity and impaired neuronal activity were rescued by specifically targeting PMI expression into neurons (Fig. 6A,C), consistent with the tissue-specific restoration of mitochondrial morphology (supplementary material Fig. S2). Although reduced ATP production by PMI mutant mitochondria probably contributes to the synaptic defects observed, we asked whether PMI inactivation could also affect the transport of mitochondria to synapses. Indeed, although drp1 mutant flies have normal ATP levels, their mitochondria are too massive to be transported to synapses, resulting in bang sensitivity and impaired neuronal activity (Fig. 6A) (Verstreken et al., 2005). We therefore imaged optic lobes that contain a regular succession of cortical and neuropil layers made up of cell bodies and synapses. Whereas mitochondria were highly enriched in the neuropil layers of controls, their density was reduced in neuropils and increased in cell bodies of PMI mutants (Fig. 6D; supplementary material Fig. S6). These phenotypes were restored by targeting PMI expression in neurons (supplementary material Fig. S6B,D). In conclusion, PMI-dependent mitochondrial morphogenesis is required for neuronal activity, efficient transport of mitochondria to synapses, and in consequence, for normal behaviour.

Our results show that the mitochondria inner-membrane protein PMI plays a central role in regulating crista morphology. Could this function be related to mechanisms previously described for other inner-membrane proteins? It was recently reported that loss-of-function mutants for Mitofilin or proteins of the MitOS/MICOS/MINOS complex are devoid of crista junctions and have closed cristae (Rabl et al., 2009; Harner et al., 2011; Hoppins et al., 2011; von der Malsburg et al., 2011). In contrast, Mitofilin overexpression causes crista branching, supporting the idea that Mitofilin might negatively bend membranes to maintain crista junction integrity (Rabl et al., 2009). None of these phenotypes were observed in PMI mutant mitochondria. Consistently, our previous data have shown that PMI associated with crista tubules and not to crista junctions (Rival et al., 2011). These data suggest that PMI is probably not a protein involved in the maintenance of crista junctions. The e and g non-catalytic subunits of F0F1 ATP synthase are two other well-characterised determinants of crista morphology. Yeast mutants for these proteins exhibit branched and even concentric (or ‘onion-like’) crista tubules, which probably result from fusion of cristae (Paumard et al., 2002; Rabl et al., 2009; Velours et al., 2009). It has been proposed that by creating membrane constraints, F0F1 e/g su shape crista tips and prevent inter-crista fusion (Rabl et al., 2009; Velours et al., 2009). As for F0F1 e/g su mutants, PMI mutant mitochondria contain highly bent cristae in adult neurons. However, although PMI mutant cristae can form hemi-concentric stacks of membrane that partly evoke the onion-like structure, they appear as a disorganised tangle of membranes that are never branched or interconnected. This suggests that PMI is regulating crista shape by a mechanism independent of F0F1 e/g su. How are the cristae maintained as straight and not bent entities? It was recently proposed that the negative and positive bending activities of Mitofilin and F0F1 e/g su, respectively, could antagonize each other to perform this function (Rabl et al., 2009). This model does not exclude the possibility that proteins whose function is to straighten the crista membranes exist. Given that loss of the PMI in adult neurons results in crista bending, it is tempting to propose that PMI could be one of these putative straightening factors. However, the analysis of cristae in PMI mutant embryo does not support this hypothesis. Although mitochondria cristae in embryos are highly elongated, they are still able to form straight and evenly spaced parallel tubules. Moreover, PMI overexpression in embryonic and in adult tissues provokes a reduction of the cristae length. These results suggest that the PMI mutant phenotype results from the over-growth of crista membranes that create hyper-elongated tubules. In this case, the phenotype observed in adult neurons would be the result of the cumulative effects of the uncontrolled crista growth during the course of development. We therefore propose that PMI controls the elongation of crista tubules by negatively regulating membrane biogenesis.

PMI mutant flies exhibit striking signs of altered oxidative metabolism including decreased respiratory activity, weaker ΔΨm, decreased ATP production and increased production of ROS. What could be the link(s) between PMI inactivation and respiratory chain dysfunction? The overall size increase and the abnormal cytoplasmic distribution of PMI mutant mitochondria are not likely to be involved since drp1 mutant mitochondria, which share similar defects, do not have altered oxidative metabolism (our data Fig. 4, and Ishihara et al., 2009; Wakabayashi et al., 2009). Since cristae concentrate respiratory chain complexes (Gilkerson et al., 2003; Vogel et al., 2006; Wurm and Jakobs, 2006), it is possible that abnormal crista morphogenesis alters the crista membrane micro-environment and in turns impairs oxidative metabolism. Consistently, in Mitofilin mutants, loss of crista junctions results in defective oxidative metabolism (John et al., 2005; Rabl et al., 2009). Similar observations were made in cells mutant for the cristae-associated proteins OPA1 and Prohibitin (Schleicher et al., 2008; Zanna et al., 2008; Osman et al., 2009a). Our finding that protein but not mRNA level of ETC subunits is decreased in PMI mutant flies supports this hypothesis. Indeed, the aberrant crista membrane organisation triggered by loss of PMI might affect membrane insertion of the respiratory chain complexes, resulting in their degradation. This, however, does not reflect a general destabilisation of the inner membrane protein pool since ATP synthase is not affected. Finally, in contrast to F0F1 e/g su mutants in which loss of mtDNA affect oxidative metabolism (Paumard et al., 2002), mtDNA content was not significantly altered in PMI mutants. These data support the idea that different defects in crista membrane organization can have specific impacts on mitochondrial function.

Loss of PMI has not only consequences on inner-membrane organisation but also on the overall shape of the mitochondrion. PMI mutant mitochondria lose their tubular shape and become spherical, a phenotype that does not result from defective fission or unbalanced fusion (Rival et al., 2011). This is consistent with the fact that although drp1 and PMI mutant mitochondria look much alike in adult neurons, with a similar global shape and size, their crista morphology is very different. This difference is even more obvious at the onset of embryogenesis. Whereas in early embryos PMI mutant mitochondria form large spheres, drp1 mutant mitochondria, although highly interconnected, remain tubular. In contrast, inhibition of opa1 induces fragmentation of mitochondrial tubules into spherical entities whose diameter remains comparable to that of wild-type mitochondria. These experiments suggest that whereas fission and fusion control the length of the mitochondrial tubules, PMI controls the diameter. We therefore propose that the excessive crista elongation observed in PMI mutant mitochondria probably triggers the increased width and therefore the spherical shape of the overall organelle. This suggests that crista length determines mitochondrial diameter, and therefore that the usual tubular shape of mitochondria can be intrinsically determined by inner-membrane morphogenesis.

Generation of PMI mutant flies and control PMI+ line is described in detailed in Rival et al. (Rival et al., 2011). Briefly, because PMI is encoded by a bicistronic locus together with the putative immune gene PGRP-LD, the genomic transgene Pac[PMI*_PGRP-LD], in which only the PMI coding sequence is mutated, was inserted into PMI knock out flies at the attp2 third chromosome landing site to generate a specific PMI mutant (w; Df(PMI_PGRP-LD ^p[w+]), Pac[PMI*_PGRP-LD]attp2). In the PMI+ line, a wild-type Pac[PMI_PGRP-LD] genomic transgene was inserted at the same chromosomal location to create a control rescue line (w;; Df(PMI_PGRP-LD ^p[w+]), Pac[PMI_PGRP-LD]attp2). To overexpress PMI we generated flies stably carrying both UAS-PMI attp40 (second chromosome) and UAS-PMI attp2 (third chromosome) transgenes that were then crossed to GAL4 flies (Rival et al., 2011). Fly stocks carrying drp1¹ and drp1² alleles were from Dr P. Verstreken, UAS-mnSOD2 from Prof. J. Phillips, Pcasper[SOD2] from Dr R. J. Mockett, UAS-opa1_IR from Dr M. Guo, and UAS-opa1 from Prof. J. Chung. UAS-mito::GFP line, sqh-mito::YFP, elav-GAL4, act5c-GAL4, hsp-FLP, FRT40A ovoD, opa1, da-GAL4 and mef2-GAL4 lines were from the Bloomington Stock Center. PMI mutant embryos were obtained from homozygous PMI mutant females being therefore devoid of both maternal and zygotic PMI expression. We generated embryos devoid of drp1 maternal contribution by crossing heat-shocked w, hsp70-FLP; FRT40A, Ovo^D/FRT40A, drp1¹; sqh-mit::YFP females to w; FRT40A, drp1¹; sqh-mit::YFP males. Drosophila stocks were maintained at 25°C on a standard cornmeal agar diet.

A dozen age-matched flies (2- to 3-days old) were anesthetized and placed in a culture vial. After 30 minutes recovery from CO₂ exposure, the vial was vortexed for 10 seconds and the number of paralysed flies was immediately determined.

Assays for longevity were performed on about 80 flies in groups of 10 in separate vials. Flies were kept at 25°C on a standard cornmeal agar diet. All tubes were kept horizontally because PMI mutant flies have the tendency to stick in the medium when experiencing seizures. Viable flies were counted and their food changed on days 1, 3 and 5 of a 7-day cycle. Log rank statistical test were performed using the MedCalc software package.

Electrophysiological recordings from flight muscles were performed with intracellular electrodes essentially as described previously (Rival et al., 2006; Dupont et al., 2012). Age-matched adult female flies were fixed with cyanoacrylate adhesive at the extremity of a needle and the dorsal thoracic cuticle was removed to expose the dorso-longitudinal flight muscle (DLM). A 10 V electrical stimulation was applied to the brain with a pair of tungsten microelectrodes isolated from the ground. This stimulated the giant fibre pathway that includes the DLM motor neurons (Tanouye and Wyman, 1980; Elkins and Ganetzky, 1990).

Primary antibodies used were: mouse anti-ATP synthase monoclonal antibody MS507 (MitoSciences; 1∶300) for mitochondria; mouse anti-Bruchpilot nc82 monoclonal antibody (Hybridoma bank; 1∶100) for synapses. Secondary antibodies: Alexa Fluor 488 goat anti-mouse IgG, Alexa Fluor 546 goat anti-mouse IgG.

For in situ detection of mit::YFP, embryos were dechorionized with bleach, glued onto a glass coverslip, and covered with oil to prevent desiccation. For in situ detection of mito::GFP, tissues were dissected in PBS, fixed for 15 minutes in 4% formaldehyde, and washed for 5 minutes. For immunolabelling, tissues were dissected in PBS, fixed for 30 minutes in 4% formaldehyde, washed for 5 minutes, permeabilised for 2 hours in 0.5% Triton X-100 in PBS, and saturated for 1 hour in 0.5% BSA, 0.1% Tween 20 in PBS (PBTB). Primary antibody diluted in PBTB was incubated overnight at 4°C, and later washed for 1 hour in 0.1% Tween 20 in PBS (PBTw). Secondary antibody was incubated for 2 hours and washed for 1 hour in PBTw. Preparations were mounted in Vectashield/DAPI (Vector). An LSM780 Zeiss confocal microscope was used for imaging.

Adult fly heads, thoraxes and embryos were fixed in 2% paraformaldehyde, 2.5% glutaraldehyde, 5 mM CaCl₂, 0.1 mM sodium cacodylate for 24 hours at 4°C. This was followed by 2-hour post-fixation in 2.5% glutaraldehyde, 0.8% osmium tetroxide, 0.1 mM sodium cacodylate at 4°C. Ultrathin Epon plastic sections were examined with a Leo 912 Zeiss microscope at 100 kV.

DHE staining was essentially performed as described previously (Owusu-Ansah et al., 2008). Tissues were dissected in PBS and incubated for 5 minutes in 30 µM DHE (in Schneider's medium) at room temperature, then washed and fixed for 8 minutes in 4% formaldehyde. DHE (excitation 350 nm, emission 430 nm) is oxidized by superoxide free radicals into ethidium (excitation 530 nm, emission 620 nm). The ethidium signal was detected with a LSM510 Meta confocal microscope (Zeiss) using a HeNe 543 nm laser, and a 560–660 nm window of detection. Pixel density was determined using ImageJ software.

Adult flight muscles (DLM) were rapidly dissected in Schneider's medium and incubated for 45 minutes, at room temperature, in 2 µM JC-1 solution (in 1% cyclodextrine Schneider's medium). Muscles mounted in Schneider's medium were immediately imaged with a LSM780 confocal microscope (Zeiss). For detection of green fluorescent monomers, and red fluorescent aggregate, excitation wavelengths were, respectively, 488 nm and 561 nm. A confocal image was taken of each prepared flight muscle, and the number of JC-1 red fluorescent aggregates automatically counted using ImageJ software.

Extracts from five adult flies were immediately deproteinised by perchloric acid precipitation (deproteinisation kit, Abcam) to eliminate ATP-consuming enzymes. ATP quantification was performed according to manufacturer's instruction (ATP fluorometric kit, Abcam). Fluorescence was measured with a Hitashi FL4500 fluorospectrophotometer (excitation/emission wavelengths: 535 nm/587 nm). Glycerol phosphate background was subtracted.

The Apoptag Red In Situ Detection Kit (Chemicon) was used to detect apoptotic nuclei in adult fly brains and flight muscles. Tissues were dissected and fixed in 4% formaldehyde in PBS. Fixed tissue were washed in PBS and apoptosis detected according to the manufacturer's instructions.

Mitochondrial respiration was assayed at 37°C by high-resolution respirometry using an OROBOROS Oxygraph. The DatLab software package (OROBOROS, Innsbruck, Austria) was used for data acquisition (2-second time intervals) and analysis, including calculation of the time derivative of the oxygen concentration, signal deconvolution dependent on the response time of the oxygen sensor and correction for instrumental background oxygen flux. State III respiration was assayed by homogenising two flies using a pestle in MiR05 respiration buffer (20 mM HEPES, 10 mM KH₂PO₄, 110 mM sucrose, 20 mM taurine, 60 mM potassium lactobionate, 0.5 mM EGTA, 3 mM MgCl₂, 1 g/l fatty-acid-free BSA) in the presence of 5–10 mM ADP. In particular, respiration buffer was supplemented with 2 mM malate plus 10 mM glutamate for complex I activity measurements and 1 mM rotenone plus 10 mM succinate for complex II activity measurements.

Protein extracts from adult flies were prepared by grinding flies in lysis buffer [25 mM Tris-HCl pH 7.4, 300 mM NaCl, 5 mM EDTA, 10% (w/v) SDS, 1 mM PMSF, 0.5% (w/v) Nonidet-40] containing the protease inhibitors leupeptin, antipain, chymostatin, pepstatin (Sigma) at the manufacturer's recommended dilution. The suspensions were cleared by centrifugation at 15,700 g for 10 minutes at 4°C and supernatants were mixed with 2× SDS loading buffer. For SDS-PAGE equivalent volumes of protein lysates were resolved on 10% Precast Gels (Invitrogen) and transferred onto nitrocellulose membranes. The membranes were blocked in TBS [0.15 M NaCl, 10 mM Tris-HCl (pH 7.5)] containing 5% (w/v) dried nonfat milk (blocking solution) for 1 hour at room temperature, and probed with the indicated primary antibody before being incubated with the appropriate HRP-conjugated secondary antibody. Antibody complexes were visualized by Pierce enhanced chemiluminescence (ECL). Primary antibodies employed in this study were OXPHOS rodent cocktail 1∶1000 (Mitosciences), ATP synthase subunit alpha 1∶1000 (Mitosciences) and α-tubulin 1∶5000 (Sigma).

Analysis of mitochondrial DNA content was performed by quantitative real-time PCR on an Mx4000 real-time cycler (Stratagene) using the QuantiTect SYBR Green PCR system (Qiagen). Specific primers for mitochondrial DNA (Fw primer: 5′-CAACCATTCATTCCAGCCTT-3′ and Rev primer: 5′-GAAAATTTTAAATGGCCGCA-3′) were designed using the primer-BLAST tool (NCBI). The relative levels of mitochondrial DNA were normalised to nuclear DNA using primers to dmActin (Dm_Act79B, QuantiTect Primer Assay, Cat. No. QT00967393). Quantification was performed using the comparative Ct method.

qRT-PCRs were performed on adult fly mRNA extracts. The following primers were used: 5′-CCAGGAAGCAGGCAAAGTAG-3′ and 5′-ATGGTCTGGGCTACGGAGAT-3′ for NDUFb8/CG3192, and 5′-CTACGAGCAGTACCGCAACA-3′ 5′-AGTCGATGATCCAGCGGTAG-3′ for SDHb/CG3283. The amount of mRNA detected was normalized to control rp49 mRNA values (5′-GACGCTTCAAGGGACAGTATCTG-3′ and 5′-AAACGCGGTTCTGCATGA-3′). Normalized data were used to quantify the relative levels of a given mRNA according to cycling threshold analysis (ΔCt).

We thank J. P. Chauvin, A. Aouane, and F. Richard (IBDML) for electron microscopy, Dr J. Devaux (CRN2M) for analysis of electrophysiological recordings, Prof. P. Belenguer and Dr L. Arnauné for helpful discussion and comments on the manuscript, and C. Nguyen for measurement and quantification on EM images.