Phenotypic rescue of a Drosophila model of mitochondrial ANT1 disease

Drosophila melanogaster is a convenient animal model for human genetic diseases, including those that are associated with mitochondrial dysfunction. Autosomal dominant progressive external ophthalmoplegia (adPEO) is a relatively common presentation of mitochondrial disease and is associated with multiple mitochondrial (mt)DNA deletions in muscle and other tissues (Suomalainen and Kaukonen, 2001). It is caused, typically, by single gene mutations in nuclear genes that are directly, or indirectly, involved in mtDNA maintenance – including those encoding the two subunits of DNA polymerase γ, POLG1 (Van Goethem et al., 2001) and POLG2 (Longley et al., 2006), the mitochondrial helicase Twinkle (C10orf2) (Spelbrink et al., 2001), the ribonucleotide reductase small subunit RRM2B (Tyynismaa et al., 2009) and the muscle-specific isoform of the adenine nucleotide translocase ANT1/SLC25A4 (Kaukonen et al., 2000). adPEO-causing mutations have been mapped to different regions of the ANT1 coding sequence (Kaukonen et al., 2000; Napoli et al., 2001; Komaki et al., 2002; Deschauer et al., 2005). Functional studies of two such mutants – A114P and V289M – in cultured myotubes revealed that, for a given mitochondrial inner membrane potential, the rate of adenine nucleotide exchange was substantially decreased, and, under some conditions, the reversed transport of ATP into mitochondria was predicted to occur readily (Kawamata et al., 2011). Earlier modeling of these mutants in yeast (Fontanesi et al., 2004) also concluded that they impair translocase activity and lead to a defect in oxidative phosphorylation (OXPHOS). ANT1 encodes the major isoform of the translocase that is found in post-mitotic tissues – such as the heart, skeletal muscle and brain – whereas three other isogenes encode isoforms of the translocase in other cells and tissues (Chevrollier et al., 2011).

AdPEO is a late-onset disorder, but ANT1 mutations are also associated with more severe early-onset diseases that show recessive inheritance. These include mitochondrial myopathy and cardiomyopathy due to complete, or almost complete, loss of translocase function (Palmieri et al., 2005), which also manifests in mice in which Ant1 has been knocked out (Graham et al., 1997; Narula et al., 2011). ANT1 deficiency has also been proposed to be a major contributor to the clinical features of Sengers syndrome, which comprises congenital cataract, hypertrophic cardiomyopathy and skeletal myopathy (Jordens et al., 2002). In this case, the effects on ANT1 are secondary because the primary mutation is in AGK, encoding acylglycerol kinase (Mayr et al., 2012), which results in mitochondrial phospholipid defects. Dysregulation of ANT1 has also been suggested to be involved in the pathogenesis of inclusion-body myositis (Barca et al., 2013) and Rett syndrome (Forlani et al., 2010).

The molecular pathology of ANT1-associated disease has been extensively studied using samples taken from affected individuals, as well as using the Ant1-knockout mouse. However, these studies have certain limitations; in human cells, the ability to recapitulate the exact in vivo physiological conditions in affected tissues – for example, the failing heart – is severely limited. As indicated above, Ant1-knockout mice exhibit dilated cardiomyopathy that is associated with a progressive decline of cardiac functions, ending in chronic heart failure (Narula et al., 2011). Conversely, Ant1 overexpression is cardioprotective (Wang et al., 2009). The knockout exhibits lactic acidosis and the hyperproliferation of abnormal mitochondria in muscle (Graham et al., 1997). The increased production of reactive oxygen species (ROS) in mitochondria that have been isolated from the affected tissues of knockout animals is also associated with mitochondrial genomic damage (Esposito et al., 1999). More surprisingly, loss of Ant1 in the brain entrains increased mitochondrial membrane potential and resistance to apoptosis, leading to the suggestion that the protein confers natural susceptibility to excitotoxicity (Lee et al., 2009). Muscle mitochondria from the knockout mouse show decreased proton leak (Brand et al., 2005), which might explain these observations, but does not necessarily elucidate the functional consequences of pathological mutations.

The Ant1-knockout mouse represents, by definition, complete loss of function, which does not appear to be the case for human ANT1 in disease, except in rare cases (Palmieri et al., 2005). However, the fact that ANT is encoded in mammals by a multigene family means that the knockout of one isoform might entrain effects on the regulation of others, which can also compensate for any defect. This has been hypothesized to account for the lack of retinal pathology in the Ant1-knockout mouse (Phillips et al., 2010). For these reasons, a more appropriate model is required that is based on actual disease-relevant mutations and in which the effects of genetic redundancy can be set aside.

In Drosophila, ANT is also encoded by a small gene family. In this case, however, the major isoform, which is encoded by the gene stress-sensitive B (sesB; FlyBase annotation CG16944), is ubiquitously expressed at a high level, whereas the mRNA for the second isogene, Ant2 (FlyBase annotation CG1683), is found only at low levels, except in testis (www.flyatlas.org). A mutant allele of sesB, sesB¹, exhibits a phenotype of susceptibility to mechanical-stress-induced seizures (‘bang sensitivity’) and developmental delay (Zhang et al., 1999), as well as physiological abnormalities in the tubule (Terhzaz et al., 2010). The combination of developmental delay and bang sensitivity is shared with other Drosophila mutants that exhibit perturbed mitochondrial OXPHOS, including alleles of knockdown (kdn), which encodes citrate synthase (Fergestad et al., 2006), and technical knockout (tko^25t), which encodes mitoribosomal protein S12 (Royden et al., 1987; Shah et al., 1997; Toivonen et al., 2001). A further bang-sensitive mutant in the gene easily shocked, which encodes ethanolamine kinase (Pavlidis et al., 1994), shares the phenotype of decreased steady-state ATP levels with sesB¹, tko^25t and some alleles of kdn (Fergestad et al., 2006) and might also act mitochondrially. tko^25t manifests multiple deficiencies of the respiratory chain and OXPHOS, and exhibits an impaired response to sound and defective male courtship (Toivonen et al., 2001).

Importantly, when the human and Drosophila sequences are aligned, point mutation L289F in sesB¹, a hydrophobic substitution in the final transmembrane-spanning helix of the protein (Pebay-Peyroula et al., 2003), is adjacent to the site of the hydrophobic substitution V298M in adPEO (Fig. 1A). Although the sesB¹ phenotype differs in its tissue specificity from that of human ANT1-associated diseases, it presents a more disease-relevant genotype than the Ant1-knockout mouse, and is effectively free of the effects that are attributable to genetic redundancy. It therefore offers a convenient tool for understanding the metabolic and cellular consequences of ANT1 deficiency, and for testing strategies to compensate for pathological defects.

In previous studies, we used transcriptomic analysis to profile changes in gene expression in the tko^25t mutant compared with wild-type flies. The results implied that there is a metabolic shift in tko^25t flies, with diversion of pyruvate to lactate to regenerate NAD⁺, increased mobilization of dietary fat and protein, and the induction of anaplerotic pathways to maintain the flow of carbon skeletons for biosynthesis (Fernández-Ayala et al., 2010). We have also shown that the tko^25t phenotype can be partially suppressed by additional dosage of the mutant gene (Kemppainen et al., 2009) or by altering mtDNA background (Chen et al., 2012), which appeared to act by increasing mitochondrial biogenesis. Therefore, we aimed to analyze metabolism and gene expression in sesB¹ flies and to test different strategies to compensate the mutant phenotype, using similar approaches as those used previously for tko^25t.

Because the sesB¹ and tko^25t phenotypes overlap, transcriptomic analysis can address what underlies the differences and similarities between the two mutants at the molecular level. These two classes of mutant – i.e. the one affecting mitochondrial protein synthesis, the other affecting intracellular nucleotide transport and synthesis – are common causes of mitochondrial disease (Saada, 2004; Sharer, 2005; Jacobs and Turnbull, 2005). Therefore, understanding both the common and the specific features of their molecular pathology should lead to more accurate diagnosis and targeted therapy.

In the mouse, transcriptomic studies have, thus far, been reported only in the skeletal muscle of 9- to 12-month-old animals for a limited set of known mitochondrially targeted gene products (Subramaniam et al., 2008), providing evidence of a paradoxical or futile upregulation of mitochondrial biogenesis. Here, we extend this to a more comprehensive analysis involving the whole genome and the whole organism in the Drosophila melanogaster model.

In order to analyze metabolism and gene expression in sesB¹ mutant flies, and to test possible strategies for suppressing the mutant phenotype, we first verified the genotype (Fig. 1A) and phenotype (Fig. 1B–H) of the available strain, which was further backcrossed into standard wild-type backgrounds (Oregon R and Canton S). Compared with tko^25t (Chen et al., 2012; Kemppainen et al., 2009), the developmental delay of sesB¹ was less pronounced, notably in males (Fig. 1B), whereas the bang sensitivity of sesB¹ adults, when tested at 3 days of age, was greater (Fig. 1C). sesB¹ flies were also deaf (Fig. 1D), as judged by a behavioral assay for response to sound, as used previously to study tko^25t flies (Toivonen et al., 2001). Despite the short lifespan (Fig. 1E), the muscle mitochondrial network in young adult flies was intact (Fig. 1F), unlike that seen in the case of other degenerative mutations that affect mitochondrial functions, such as Pink1⁵ (Deng et al., 2008). Female flies were essentially sterile. At 2 days of age, ovary size was, typically, decreased (Fig. 1G), although eggs were eventually produced in normal, or near normal, numbers. However, eggs that were laid developed only as far as the late-embryo stage (Fig. 1H), with a tiny number of survivors reaching adulthood (~0.1% of the number of progeny from control crosses). Males were fertile but, in courtship assays, they displayed a defect similar to that observed previously for tko^25t (Toivonen et al., 2001): in two separate experiments, only one of 11 and one of 13 individual sesB¹ males reached copulation with single wild-type females, whereas ten out of ten wild-type males successfully copulated.

In order to test how the sesB¹ mutation affects the properties of the ANT protein, we measured the rate of respiration of isolated mitochondria from sesB¹ and control flies in the presence of different substrates, with and without the addition of 1 mM ADP. The respiratory control ratio (RCR), computed as the respiration rate in the presence of ADP divided by that in its absence, represents the degree to which respiration is limited by ATP synthesis, which is in turn dependent on ANT activity. On both complex-I- and complex-III-linked substrate mixes, mitochondria from sesB¹ flies showed a significant drop in the apparent RCR (Table 1; Fig. 2A), consistent with decreased ANT activity limiting both ATP synthesis and respiration. Note, however, that because of the secondary activation of isocitrate dehydrogenase by ADP, the complex-I-linked data is not a true reflection of the RCR (Miwa et al., 2003). There was a 50% decrease in RCR, even using the complex-III-linked substrate glycerol-3-phosphate.

Steady-state ATP levels were also consistently lower in sesB¹ flies than control flies (Fig. 2B). Under conditions where respiration is limiting, re-oxidation of NADH should depend on cytosolic enzymes, such as lactate dehydrogenase. This was reflected here by increased levels of lactate (Fig. 2C). In order to investigate the effects of the sesB¹ mutation on the respiratory chain, we analyzed the activity of complex I and IV in mitochondrial extracts from sesB¹ and control flies, using blue-native electrophoresis (BNE) combined with histochemical staining (Fig. 2D; supplementary material Fig. S1). Complex I activity in sesB¹ was indistinguishable from that in wild-type flies, but there was a clear decrease in complex IV activity, especially in females.

We implemented a gene expression analysis of sesB¹ flies using procedures identical to those used previously to analyze tko^25t. To minimize strain-specific effects, the flies analyzed using the Affymetrix Drosophila 2.0 microarray were, as in the case of tko^25t, an F1 hybrid of sesB¹ that had been backcrossed to Canton S or Oregon R backgrounds, and the results were compared with those of a corresponding control F1. The changes in the expression of a subset of genes were then verified by quantitative RT-PCR (Table 2). Analyses were conducted separately for males and females, in order to clarify the bisexual versus sex-specific aspects of the mutant phenotype.

After statistical filtering, we identified 903 probesets that detected genes whose expression in sesB¹ females differed at least twofold from that of wild-type controls (supplementary material Table S1), and 366 such probesets in males (supplementary material Table S2). Only 181 probesets were shared between the sexes (Fig. 3A; supplementary material Table S3), 177 of which showed directionally congruent changes, most of which were quantitatively very similar, with eight times as many genes that were upregulated as downregulated. Gene ontology (GO) analysis revealed that the genes represented by the 181 bisexually-regulated probesets were mostly annotated with molecular functions in metabolism (Fig. 3E). Of the probesets that were regulated in a sex-specific manner, downregulation was slightly more prevalent than upregulation. We inspected in more detail the expression patterns and GO terms of the 20 most up- or downregulated sex-specific annotated genes in each sex (Table 3), which revealed a consistent pattern. The sex-specific upregulated genes fell mostly into similar categories as those regulated bisexually, suggesting that these represent pathways regulated in a similar way in the two sexes, but in which specific genes were rejected in one sex. This might have been due to the stringent statistical filtering, or, alternatively, to the quantitative limitations of the method, in comparison with quantitative PCR (see Table 2). Some of the genes that were specifically downregulated in males were also in these categories. However, the remaining genes that were downregulated in males, and almost all of those that were downregulated in females, were associated with reproduction and broadly consistent with the organismal phenotype. Thus, the genes that were most downregulated in females were predominantly involved in oogenesis – including chorion formation and cell division functions – which is consistent with defective oocyte development. In males, the most prominent set of sex-specific downregulated genes were expressed primarily, or exclusively, in the accessory glands and have been implicated in pheromone perception and other aspects of courtship behavior, in which sesB¹ males were defective. Global GO analysis of the annotated genes that showed alterations in their expression in a sex-specific manner in sesB¹ flies (supplementary material Tables S4, S5) was consistent with this ‘top 20’ analysis, although many ovary-specific or ovary-enriched genes that are involved in oogenesis are listed under other categories – such as anatomical structure development, cell cycle, cell differentiation, cell division, chromosome or cytoskeleton organization, mitosis, chromosome segregation or DNA metabolic process.

Next, we compared the patterns of change in gene expression in sesB¹ flies with those seen in tko^25t (Fig. 3B–D; supplementary material Tables S6–S10) in order to reveal both common and distinct pathways that are affected in these models of two major classes of mitochondrial disease. For coherence with the previous analysis of tko^25t (Fernández-Ayala et al., 2010), we applied a slightly more relaxed criterion of a statistically robust expression difference of ≥1.5-fold. In males, approximately two-thirds of the probesets that showed altered expression in tko^25t were also altered in sesB¹, and half of the genes that showed altered expression in sesB¹ were also altered in tko^25t with, once again, all but a handful of genes showing the same directional difference in their expression, most of which were quantitatively very similar (Fig. 3B; supplementary material Table S7). In females, the vast majority of the probesets that showed altered expression in tko^25t were also altered in sesB¹ flies (383 of 473 probesets), whereas a larger number of the genes that showed altered expression in sesB¹ females (611 probesets) did not appear in the tko^25t list (Fig. 3C; supplementary material Table S6). Almost all of the genes that showed the greatest changes in expression in tko^25t flies, whether regulated bisexually or in a sex-specific manner, were altered to a similar degree in sesB¹ (supplementary material Table S10). Of those showing the greatest sex-specific changes in sesB¹ females, as listed in Table 2, 15 out of 20 of those that were upregulated were also upregulated in tko^25t, whereas only seven out of the 20 most downregulated genes in sesB¹ females were also downregulated in tko^25t females. Considering both mutant backgrounds, 126 probesets were regulated in a similar manner in both sexes in both mutant strains (Fig. 3D; supplementary material Table S8), whereas a larger number were sex-specific (Fig. 3D; supplementary material Table S9). GO analysis of the genes that were up- or downregulated in sesB¹ females but not in tko^25t females placed these genes in diverse categories. However, more than half of the 122 annotated genes that were downregulated (but almost none of the upregulated genes) are expressed uniquely, or predominantly, in the female reproductive system (www.flyatlas.org).

The observed changes in the expression of metabolic genes in sesB¹ can be broadly rationalized, as in the case of tko^25t, in the following manner. First, the bulk of ATP production is shifted to glycolysis, with conversion of pyruvate to lactate to regenerate NAD⁺. Second, fatty acids and amino acids are mobilized, and anaplerotic pathways induced to maintain the flow of carbon skeletons for biosynthesis. Thus, there was upregulation of a set of enzymes that converged on pyruvate and acetyl-CoA in both sexes – fructose-1,6-bisphosphatase, aldolase, phosphoenolpyruvate carboxykinase (Pepck), lactate dehydrogenase, acetyl-CoA synthetase and aldehyde dehydrogenase, as well as alcohol dehydrogenase in females (supplementary material Table S11). A second Pepck isogene (CG10924) and a testis-specific lactate dehydrogenase gene were downregulated (in males). There was also an overlap between the sets of genes that showed altered expression in sesB¹ and those that are altered under starvation (Farhadian et al., 2012), although some of the changes were in opposite directions (supplementary material Table S12).

In previous work, we have described the partial alleviation of the phenotypes of Drosophila mutants that represented mitochondrial disease models associated with cytochrome oxidase deficiency (Fernández-Ayala et al., 2009; Kemppainen et al., 2014), complex I deficiency (Sanz et al., 2010a), DNA polymerase γ disease (Humphrey et al., 2012) and disorders of mitochondrial protein synthesis (Kemppainen et al., 2009; Chen et al., 2012). The similar organismal phenotypes and the overlapping patterns of altered gene expression in tko^25t and sesB¹ led us to question whether the same strategies might be effective in sesB¹ flies.

We have previously reported that the BER-1 mtDNA background, which produced a clear phenotypic improvement in tko^25t, had no effect on the developmental phenotype of sesB¹, although there was a partial rescue of its short lifespan (Chen et al., 2012). Note, however, that observations of eclosion timing and bang sensitivity that we reported in that publication were made after backcrossing sesB¹ directly into the BER-1 cytoplasmic background, which remained infected with the endosymbiont Wolbachia. Furthermore, the continued presence of Wolbachia compromised the increase in the lifespan of sesB¹ that was conferred by the BER-1 mtDNA background (see figure 3 in the supplementary data of Chen et al., 2012). Therefore, we investigated whether the BER-1 mtDNA background, in the absence of Wolbachia, was able to rescue the developmental phenotype of sesB¹.

BER-1 females, after removal of Wolbachia by tetracycline treatment (Dobson et al., 2002; Chen et al., 2012), were backcrossed for more than five generations to sesB¹ in order to create a cybrid strain that contained BER-1 mtDNA but the original nuclear genome of sesB¹. Phenotypic testing of the cybrid flies indicated a partial rescue of both developmental delay (Fig. 4A) and bang sensitivity (Fig. 4C). In previous work, we found that the partial rescue of tko^25t by the BER-1 mtDNA background was accompanied by an increased expression level of the Drosophila PGC-1α homolog spargel (Chen et al., 2012) and, furthermore, that overexpression of spargel was able to phenocopy the effect. However, in the sesB¹ flies we did not find any effect of the BER-1 mtDNA background on the expression level of spargel (Fig. 4D), indicating that this is not the mechanism of rescue.

We also tested directly whether spargel overexpression was able to suppress the sesB¹ phenotype. Despite an apparent partial rescue of bang sensitivity (Fig. 4C), there was no rescue of developmental delay (Fig. 4B), despite the increase in spargel expression at the RNA level that was conferred by the extra copy of the gene (Fig. 4D). spargel overexpression produced only a minimal rescue of the sterility of sesB¹ females when mated to wild-type males (Fig. 4D), which was also the case for the BER-1 flies, with only a slight increase in the number of survivors as compared with that resulting from control crosses.

Use of a respiratory chain bypass for complexes III and IV, via the alternative oxidase (AOX) from the sea squirt Ciona intestinalis, should facilitate electron flow through complex I, if the latter is inhibited by a limitation on ATP synthesis imposed by ANT dysfunction. This is predictable from the fact that complex I, combined with AOX, pumps fewer protons per electron than the regular respiratory chain. Any additional restriction due to a decrease in complex IV activity (Fig. 2D) might also be circumvented. Furthermore, respiration through complex II or other dehydrogenases that donate electrons to ubiquinone should also be facilitated under such conditions, because, when combined with AOX, they do not pump any protons. Thus, AOX expression should relieve sesB¹ phenotypes that result from disturbed redox or metabolic homeostasis, but not those caused by decreased ATP synthesis per se.

Testing showed that AOX expression was unable to rescue developmental delay (Fig. 5A), although there was, once again, a minor degree of rescue of bang sensitivity (Fig. 5B), which was not further enhanced by da-GAL4-driven high-level expression. Expression of the alternative mitochondrial NADH dehydrogenase Ndi1 from yeast (Fig. 6A) was also unable to rescue developmental delay and, instead, tended to increase it.

To test for the possible rescue of female sterility, we crossed sesB¹ females (1- to 4-day-old virgins) that also carried transgenic upstream activator sequence (UAS)-AOX (Fernández-Ayala et al., 2009), with or without the ubiquitously acting da-GAL4 driver, with wild-type males. Control females that were heterozygous for sesB¹ gave substantial numbers of progeny (Fig. 5C), but, unlike non-transgenic sesB¹ homozygous females or the flies that had been tested previously in the BER-1 mitochondrial or spargel overexpressor backgrounds, those flies also endowed with UAS-AOX gave progeny well in excess of the ~0.1% survivor rate obtained routinely from sesB¹ females. In the presence of only the transgene encoding UAS-AOX, the number of progeny was ~3% of the number from control crosses, but the additional presence of the da-GAL4 driver raised this to 15–20% (Fig. 5C). Ndi1 expression failed to rescue female sterility, with only a tiny number of progeny (Fig. 6B). Note that the experiments using the transgenes that encoded UAS-AOX and UAS-Ndi1 were performed in the same genetic background, thus the differences are biologically meaningful, as well as statistically significant.

The short lifespan of sesB¹ flies was marginally improved by AOX expression – AOX-expressing males showed a statistically significant increase in their mean lifespan of 3–4 days compared with that of control groups (Fig. 5D), but the effect in females was negligible (Fig. 6B). However, given that the rescue was quantitatively minor and sex-specific, a contribution of the genetic background cannot be excluded.

In this study, we characterized the biochemical and phenotypic consequences of the sesB¹ mutation in the major ANT gene of Drosophila, and tested different strategies to complement its phenotype in vivo. We determined that the mutation limits respiration, causes a moderate deficiency of ATP and increases the level of lactate. Most aspects of the adult phenotype, and alterations in gene expression that affect metabolism, are shared with the tko^25t mutant, revealing common underlying molecular pathways. However, sesB¹ flies also exhibit the distinct, and sex-specific, phenotype of female sterility, which is associated with a downregulation of a broad range of genes that are associated with oogenesis. Female sterility was overcome by the expression of AOX from Ciona, but this did not rescue developmental delay and only partially alleviated bang sensitivity. Crossing sesB¹ into the BER-1 mtDNA background produced modest apparent improvements in the developmental phenotype but had little impact on female fertility, whereas the effects of spargel overexpression were even more minor.

The sesB¹ biochemical phenotype is the same as that which would be predicted if the mutation causes a decreased rate of adenine nucleotide exchange. The RCR can be construed as reflecting the degree of functional coupling between ATP synthesis and respiratory electron flow. A low RCR (Fig. 1A), combined with ATP deficiency (Fergestad et al., 2006; Terhzaz et al., 2010), as verified here (Fig. 1B), thus reflects either increased proton leak across the inner mitochondrial membrane and/or a limitation on ATP synthesis, either of which are plausible molecular consequences of the sesB¹ mutation, which affects the final transmembrane segment of the protein. Despite modeling in yeast, which supported uncoupling as the mechanism of adPEO mutations in ANT1 (Wang et al., 2008), we favor the latter explanation because it is more easily reconciled with the fact that the mutation is recessive and is parsimonious with the reported biochemical phenotype of the V298M adPEO mutation (Kawamata et al., 2011), which affects the residue adjacent to that mutated in sesB¹. Previously, it was shown that decreased expression of wild-type sesB decreased the net proton conductance of the inner mitochondrial membrane, whereas its overexpression increased it (Brand et al., 2005). Thus, increased proton leak seems an unlikely explanation for the effect of the sesB¹ mutation. Furthermore, the apparent decrease in the amount of functional cytochrome oxidase (Fig. 1D) is a plausible response to a limitation of ATP synthesis as a mechanism to restrict electron flow and prevent increased membrane potential that would result in overproduction of ROS, as seen in the mouse Ant1 knockout (Esposito et al., 1999). Increased proton leak would be expected to produce, if anything, an opposite response to that observed here.

Based on the findings from transcriptomics and metabolite analysis, we found evidence of a remodeling of metabolism that should favor glycolytic ATP generation – pyruvate was converted to lactate to regenerate NAD⁺, fatty acid and amino acid mobilization and catabolism were induced, as well as the induction of anaplerotic pathways to compensate for a presumed decrease in the mitochondrial import of pyruvate. Furthermore, the major stress-response gene, Hsp22, a mitochondrial chaperone of the lens-crystallin superfamily, was upregulated (Morrow et al., 2000; Michaud et al., 2002). As a programmed response to transient mitochondrial insufficiency, we hypothesize that this is sufficient to sustain metabolism and support development, even if the outcome is sub-optimal, yielding flies that are bang-sensitive and defective in reproduction.

The sesB¹ phenotype has many similarities with that of tko^25t, including decreased ATP levels, developmental delay, bang sensitivity, male courtship defect, hearing impairment and a similar remodeling of the metabolic component of gene expression (Toivonen et al., 2001; Fernández-Ayala et al., 2010; Chen et al., 2012). When considering potential future treatments of these distinct classes of mitochondrial disease, such common molecular pathways might prove fruitful targets.

One major difference between the two mutants is that tko^25t has an essentially normal lifespan (Chen et al., 2012), whereas that of sesB¹ is drastically shortened (Fig. 1D). The adult fly is essentially post-mitotic; therefore, in contrast with development – when active cell growth and cell division take place – mitochondrial protein synthesis in adults is required only to keep pace with protein and organelle turnover. A greatly decreased capacity for mitochondrial protein synthesis, as in tko^25t, might thus have severe effects on development, yet still be above the rate that is required for turnover in the adult. If this is the case, the capacity for OXPHOS should gradually increase in tko^25t adults, eventually reaching a level similar to that of wild-type flies, with only a minimal, and probably undetectable, effect on aging.

By contrast, the limitation on respiration and OXPHOS that is conferred by decreased ANT activity in sesB¹ should persist in the adult, imposing chronic stress and probably accounting for the drastically curtailed lifespan. One obvious potential stressor that is already associated both with ANT dysfunction and lifespan regulation is excess ROS production. AOX decreases mitochondrial ROS production, even under non-inhibited conditions (Fernández-Ayala et al., 2009; Sanz et al., 2010b), although this is not understood mechanistically. The modest lifespan extension that is produced by AOX, at least in sesB¹ males, could thus be mediated through an effect on ROS. However, this increase is quantitatively too small to exclude an effect of genetic background.

A second major difference between sesB¹ and tko^25t is that tko^25t females are fertile, despite a reduced rate of egg laying (Fernández-Ayala et al., 2010), whereas sesB¹ females are sterile, giving rise to eggs that can be fertilized but fail to complete embryogenesis. This is also reflected in decreased ovary size (Fig. 1G) and the downregulation of many genes that are connected with oogenesis in sesB¹ females (Table 3; supplementary material Table S5). Although the number of survivors was slightly increased in the BER-1 mtDNA background or upon increased spargel expression, only the expression of AOX, driven by da-GAL4, was able to produce a pronounced rescue. By contrast, expression of Ndi1, which bypasses (and also alleviates ROS production at) complex I (Sanz et al., 2010a), had only a negligible effect on female sterility (and lifespan), excluding background effects from the da-GAL4 driver or the w¹¹¹⁸ recipient that was used to generate the transgenic lines. AOX was partially effective at alleviating female sterility, even when expressed only at a low level (without the da-GAL4 driver), which was also the case for its partial rescue of bang sensitivity, reminiscent of previous observations on its rescue of the locomotor defect that is caused by deficiency of dj-1β (Fernández-Ayala et al., 2009).

The fact that AOX partially rescued fertility, whereas Ndi1 did not, can be interpreted in several alternative ways. The first possibility is that rescue by AOX involves the mobilization of substrates that supply electrons downstream of complex I (such as succinate or glycerol-3-phosphate). AOX can transfer these electrons to oxygen without net proton pumping, circumventing the limitation on electron flow that is imposed by the coupling mechanism caused by ANT insufficiency. The second is that the observed downregulation of complex IV in sesB¹ flies limits any facilitation of electron flow by Ndi1. AOX bypasses complex IV and so should be unaffected. The third possibility invokes a different metabolic effect of AOX, such as decreased ROS production.

Because the standard UAS-GAL4 system is inactive in the female germline (Rørth, 1998), the AOX-mediated rescue of sterility must be attributable to its expression in somatic tissues, or in the embryo itself. Because sesB¹ females are sterile when mated to wild-type males, it is evident that zygotic expression of even sesB itself cannot rescue sterility. Thus, it seems unlikely that AOX could do so. Instead, we favor the idea that insufficiency of mitochondrial adenine nucleotide transport in one or more somatic cell types in the ovary leads to a dysregulation of oogenesis that can be rescued by AOX. The fact that so many ovary-specific genes are downregulated in sesB¹ is consistent with a general failure of oogenesis rather than a specific metabolic abnormality in developing oocytes.

The somatic cells of the ovary, in particular the follicle cells, play key roles in the patterning of the developing oocyte (Deng and Bownes, 1998; Poulton and Deng, 2007), and many molecular signaling pathways are involved (Roth et al., 1995; Neuman-Silberberg and Schüpbach, 1996; Rørth et al., 2000; Yamamoto et al., 2007; Stein et al., 2013; Fregoso Lomas et al., 2013), some of which are functionally conserved in vertebrates (Charlier et al., 2012). Moreover, cell proliferation and migration – crucial to the developmental function of the follicular epithelium – are commonly regulated by metabolite or redox signaling (Veal and Day, 2011; Hurd et al., 2012). ANT dysfunction might interfere with this signaling, whereas AOX might restore it.

ANT is an abundant component of the protein-rich inner mitochondrial membrane, probably reflecting a quantitative requirement for a high rate of adenine nucleotide transport to match the capacity of ATP synthase. For this reason, a decrease in ANT activity is likely to become limiting for ATP synthesis, with a small margin for compensation by increasing the expression or stabilization of the protein. However, globally increased mitochondrial biogenesis mediated by, for example, PGC-1α (spargel) would be a plausible compensation strategy. In previous studies, this pathway has been implicated in the partial rescue of tko^25t by the BER-1 mtDNA background (Chen et al., 2012). However, the modest phenotypic improvement in sesB¹ that was brought about by BER-1 mtDNA was not accompanied by increased spargel expression. Moreover, a twofold overexpression of spargel had no effect on the developmental delay of sesB¹ flies, indicating that a different mechanism underlies the effects of BER-1-mediated rescue.

Compared with control strains, BER-1 mtDNA contains a number of sequence polymorphisms, which affect both protein- and RNA-encoding genes (Chen et al., 2012). These might act in a variety of ways to modify the sesB¹ phenotype – e.g. by altering the functional properties of the OXPHOS system, the architecture of the inner mitochondrial membrane or the output of mitochondrial protein synthesis. An effect on mitochondrial biogenesis that is independent of the spargel pathway can also not be excluded. In order to understand how BER-1 influences sesB¹ it will be useful to compare in detail the properties of isolated mitochondria from sesB¹ and control flies containing BER-1 versus control-strain mtDNA. Regardless of the precise mechanism, our findings provide support for the idea that the mitochondrial genotype can act as a modifier in humans with recessive ANT1-linked disease (Strauss et al., 2013).

Despite its ability to rescue female sterility, AOX expression produced no improvement in the developmental delay of sesB¹ flies, consistent with the notion that this phenotype is, most probably, attributable to the limitation on ATP production that is imposed by sesB¹. ATP deficiency, specifically at synapses, has also been suggested to underlie the phenotype of a second mutant allele of sesB, sesB^org, which confers temperature-sensitive paralysis (Rikhy et al., 2003).

Overall, our findings indicate that there are at least two pathways by which ANT deficiency can lead to pathological tissue dysfunction, and the sesB¹ model now allows these to be explored in further detail. The first pathway is through bioenergetic insufficiency, manifesting as an ATP deficit, of which developmental delay is the principal indication. Its apparent alleviation by the BER-1 mtDNA background might have substantial implications for therapy, if the mechanism can be elucidated. Conversely, despite a large interest in therapeutically boosting PGC1-α expression (Wenz et al., 2008; Srivastava et al., 2009; Dillon et al., 2012; Chaturvedi and Beal, 2013; Komen and Thorburn, 2014; Noe et al., 2013; Hofer et al., 2014), our results indicate that modulating PGC1-α activity might not be a viable strategy for alleviating the features of ANT1 diseases that are associated with bioenergy deficit. Nevertheless, much stronger activation of the pathway, such as using UAS-spargel, might yet be worth testing.

The second pathway by which ANT deficiency produces a pathological outcome is the one alleviated by AOX, involving disturbed metabolic or redox homeostasis.

Ubiquitous AOX expression in both the fly and the mouse is essentially benign (Fernández-Ayala et al., 2009; El-Khoury et al., 2013). The benefit of this in future therapy (El-Khoury et al., 2014) is already indicated by the ability of AOX to alleviate the pathological defects that are caused by functional loss of complex IV (Fernández-Ayala et al., 2009; Kemppainen et al., 2014) or other mitochondrial components that are needed for OXPHOS, such as DNA polymerase γ (Humphrey et al., 2012). Here, we extend this range of targets to include ANT, but with an important caveat. Clearly, only some of the phenotypes that are associated with pathological ANT dysfunction can be mitigated by AOX. Thus, any potential use of AOX in therapy for ANT-related, and other mitochondrial, disorders should be based on a clear understanding of the underlying pathological mechanism and the metabolic status. These might differ systematically between tissues, according to nutritional state or under influences of the genetic background.

Drosophila melanogaster lines were as described previously (Toivonen et al., 2001; Fernández-Ayala et al., 2009; Sanz et al., 2010a; Kemppainen et al., 2009; Chen et al., 2012) or from stock centers (sesB¹, bss, sda, Mef2-GAL4, mito-HA-GFP). The srl^GR line (Tiefenböck et al., 2010), containing an additional genomic copy of the spargel gene, was a kind gift from Christian Frei (ETH Zürich, Switzerland). Flies were maintained at 25°C on standard medium with supplements, as described previously (Fernández-Ayala et al., 2009) except where indicated. Time to eclosion, bang sensitivity (at 25°C), response to sound and male courtship were measured as described previously (Toivonen et al., 2001). Lifespan at 25°C was determined as described previously (Sanz et al., 2010a).

The dissected ovaries from 2-day-old virgin females were placed on a slide in a droplet of PBS, and images were captured with and without coverslips by using a Nikon SMZ745T microscope with DS-Vi1 camera. The embryos were collected from mating chambers, dechorionated, placed on two-sided tape that was attached the bottom of a Petri dish and then minimally overlaid with dissection medium (137 mM NaCl, 5.4 mM KCl, 0.17 mM NaH₂PO₄, 0.22 mM KH₂PO₄, 33.3 mM glucose, 43.8 mM sucrose and 9.9 mM HEPES, pH 7.4). Embryos were imaged as described above. Indirect flight muscle mitochondria of flies that expressed GFP that was targeted to the mitochondria were imaged by using a Nikon Eclipse-Ti microscope (Nikon Instruments, Melville, NY) plus an Andor Neo sCMOS camera (Oxford Instruments, Abingdon, UK) with a 40× objective. The flies were grown for 2 days, after which the thoraxes were fixed with phosphate buffered 10% formalin pH 7.0 (FF-Chemicals, Haukipudas, Finland) for 30 minutes, washed with PBS three times for 10 minutes, after which further dissection of muscle fibers was performed with a 27-gauge needle attached to a 1 ml syringe.

ATP levels were assayed using the ATP Determination Kit (Life Technologies, Carlsbad, CA) according to the manufacturer’s instructions. For each sample, a set of ten flies were homogenized in 100 μl of 6 M guanidine hydrochloride and then centrifuged at 12,000 g_max for 5 minutes at 4°C. Supernatants were diluted tenfold in TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0), and 10 μl of the diluted sample was mixed with the reaction solution. Different genotypes were processed in biological triplicates and luminescence was measured using a Chameleon plate reader (Hidex, Turku, Finland).

Lactate was assayed by using a BioVision Lactate Colorimetric/Fluorometric Assay Kit (BioVision, Milpitas, CA) according to the manufacturer’s instructions. Samples of 30 flies per genotype were homogenized in 300 μl of lactate assay buffer and centrifuged at 12,000 g_max for 5 minutes at 4°C. Each genotype was analyzed in triplicate and sample fluorescence was recorded using the Hidex Chameleon plate reader.

Mitochondria were isolated for oxygraphic analysis and blue native electrophoresis from ~200 flies per genotype, as previously described (Miwa et al., 2003). Protein concentrations were determined by the Bradford method (Bradford, 1976). Oxygen consumption of the isolated mitochondria was measured in a Clark-type electrode chamber (Hansatech Oxyterm system) in the presence of different substrate mixes (5 mM sodium pyruvate plus 5 mM proline for complex-I-linked respiration, 20 mM sn-glycerol 3-phosphate for complex-III-linked respiration) before and after the addition of 1 mM ADP, as described previously (Fernández-Ayala et al., 2009). The oxygen consumption in the presence of the different substrate mixes was then used to compute the RCR. Blue native electrophoresis and histochemical staining for complex I and IV activity were performed on mitochondrial extracts, as described previously (Sanz et al., 2010a).

RNA extraction, microarray analysis using the GeneChip^® Affymetrix Drosophila 2.0 array (Affymetrix, Santa Clara, CA) and quantitative RT-PCR were performed as described previously (Fernández-Ayala et al., 2009; Fernández-Ayala et al., 2010). RNA isolations were performed in triplicate from batches of 30 2-day-old males and virgin females. Raw microarray data was normalized and analyzed using Genespring GX software (Agilent Technologies, Santa Clara, CA). Previous raw data from tko^25t mutant flies (Fernández-Ayala et al., 2010) was incorporated for comparative analysis, for which data normalization and analysis was performed afresh, grouping all values for the control samples. Genes that showed altered expression were clustered according to GO classes, using GO Term Mapper (Lewis-Sigler Institute for Integrative Genomics, Princeton University, NJ). For quantitative RT-PCR, cDNA was synthesized using the SuperScript^® kit (Life Technologies, Carlsbad, CA) or, for spargel RNA, using a High Capacity cDNA Reverse Transcription kit (Applied Biosystems). Analysis used a StepOnePlus instrument (Life Technologies, Carlsbad, CA) with the manufacturer’s SYBR^® Green PCR reagents and customized primers, as listed in supplementary material Table S13. Normalization to RpL32 RNA was performed as described previously (Fernández-Ayala et al., 2009).

For sequencing, genomic DNA was prepared from batches of ten to 30 adult flies, as described previously (Fernández-Ayala et al., 2009). A 500 bp fragment of the sesB gene was amplified using primer pair sesB1F: 5′-GAGGAAGGAAGGGTTCTTGC-3′ and sesB1R: 5′-CATCGTGTCCTATCCCTTCG-3′. The purified PCR products were sequenced using the same primers (Life Technologies, 3130x1 Genetic Analyzer).

In all experiments, the means were compared, as appropriate, by Student’s t-test or ANOVA, with Bonferroni correction for multiple comparisons.

We thank Niels Krol for assistance with quantitative RT-PCR analyses, Eveliina Kaulio and Outi Kurronen for technical assistance, and Laurie Kaguni for useful discussions.