clueless, a conserved Drosophila gene required for mitochondrial subcellular localization, interacts genetically with parkin

Parkinson’s disease results from the catastrophic loss of dopaminergic neurons within the substantia nigra of the brain. Significant declines in mitochondrial function are now recognized as a frequent feature of both heritable and sporadic Parkinson’s disease (reviewed by Abou-Sleiman et al., 2006). Mutations or environmental factors associated with the disease are thought to increase oxidative stress within susceptible neurons by a wide variety of mechanisms, leading eventually to mitochondrial failure, accelerated cell death and the onset of symptoms. Mutations in parkin (also known as PARK2) (Kitada et al., 1998) are found in about 50% of heritable cases of autosomal recessive juvenile Parkinsonism (Betarbet et al., 2005), and the protein is found within lesions in the affected neurons of patients with sporadic disease (Schlossmacher et al., 2002). Although less frequent, PINK1 mutations (Valente et al., 2004) are also responsible for a significant number of cases of familial Parkinson’s disease.

Studies of the Drosophila parkin (park) and pink1 orthologs (Greene et al., 2003; Wang et al., 2006) (reviewed by Pallanck and Greenamyre, 2006) strongly support the view that mitochondrial defects are central to Parkinson’s etiology. Flies that are mutant for park contain defective mitochondria in many tissues and prematurely lose dopaminergic neurons in the brain. Park is predicted to function as an E3 ubiquitin ligase and might act to remove damaged mitochondrial proteins through ubiquitylation and degradation in the proteasome. Pink1, a mitochondrially targeted serine/threonine kinase, is also required for normal mitochondrial morphology and function, possibly by facilitating mitochondrial fission/fusion (Clark et al., 2006; Wang et al., 2006; Yang et al., 2008). Overexpression of Park can rescue pink1 mutant flies, suggesting that these genes function in a common pathway (Clark et al., 2006; Park et al., 2006). Interactions of pink1 and park with genes controlling mitochondrial dynamics suggest that the mitochondrial fission pathway may be involved (Poole et al., 2008). However, the specific molecular mechanisms connecting Park and Pink1 to mitochondrial function remain imperfectly understood.

The Dictyostelium cluA gene encodes a highly conserved protein that affects mitochondrial localization within cells, but its relationship to other mitochondrial pathways is unclear (Zhu et al., 1997). Mitochondria aggregate into a single cluster in cluA mutant cells, and similar clustering occurs in Baker’s yeast (Saccharomyces cerevisiae) that is mutant for clu1, an orthologous gene that can functionally substitute for cluA (Fields et al., 1998; Dimmer et al., 2002). Metazoan eukaryotes contain CluA orthologs, such as friendly mitochondria (fmt) in Arabidopsis, a gene that is also required to prevent mitochondria from clustering (Logan et al., 2003). The C. elegans gene clu-1 and a human ortholog, KIAA0664 (‘human Clu’), exist but little information is currently available on their function.

Drosophila oogenesis (for a review, see Spradling, 1993) represents a highly favorable system for studying the functional importance of mitochondrial subcellular localization. Developing follicles grow extensively, facilitating the visualization of organelles within the oocyte and its 15 interconnected nurse cells. Moreover, mitochondria undergo a series of developmentally regulated behaviors as germ cells progress from stem cell to completed egg (Cox and Spradling, 2003). These include programmed fission after the stem cell stage, Balbiani body formation at the time of follicle formation, and extensive replication within growing follicles. Microtubule-based transport mediated by Dynein (Dhc), Kinesin (Khc) and the adaptor protein Milton (Milt) plays a crucial role in these events (Cox and Spradling, 2006). By studying how mitochondria redistribute in ovaries bearing Dhc, Khc and milt mutations, the predominant orientation of the microtubules used to transport mitochondria could be deduced (Cox and Spradling, 2006).

Here, we describe the gene ‘clueless (clu)’ encoding a Drosophila CluA ortholog that is 53% identical to human Clu, and show that clu mutations cause mitochondrial dysfunction and clustering. Similar to parkin mutants, clu null mutant adults display a shortened life span, male and female sterility, as well as severe mitochondrial abnormalities in flight muscle. Furthermore, park mutations cause clustering of mitochondria in follicle cells and nurse cells during oogenesis. clu interacts genetically with park because park/+; clu/+ trans-heterozygotes show enhanced mitochondrial clustering. Based on these findings, we propose that Clu and Park participate in a physiological and oxidative damage control pathway(s) that links mitochondrial localization and function.

The Drosophila gene CG8443 produces a single annotated transcript encoding a 1448-amino acid (aa) predicted protein that is related closely to CluA from Dictyostelium and its orthologs (Fig. 1A,B). Consequently, we named this gene clueless (clu). Clu shows 53% identity to human Clu (KIAA0664) across the entire protein. Both proteins contain a region of tetratricopeptide repeats (TPR), as well as a putative ‘Clu domain’ between amino acids 424–666 of Drosophila Clu, that is even more strongly conserved in metazoan Clu proteins (85% identity between Drosophila Clu and KIAA0664) (supplementary material Fig. S1). N- and C-termini protein segments of Drosophila Clu were expressed and used to raise antibodies in guinea pigs (Fig. 1B) (see Methods). Both antibodies recognize a single band of apparent molecular weight 160 kD on western blots (Fig. 1C) whose abundance is reduced in clu mutants (see below).

To analyze clu function, we identified multiple strains bearing single insertions located within the clu transcription unit (Table 1; Fig. 1A). Flies that are homozygous for clu^d08713 or clu^f04554 lack detectable Clu protein on western blots, whereas clu^d00713 or clu^EP0969 homozygotes contain reduced amounts (Fig. 1C). These clu mutant flies showed striking defects. Putative null clu^d08713 (Fig. 2A) or clu^f04554 homozygous flies (data not shown) are small, uncoordinated, sterile and live for only 3–7 days. They move slowly, do not fly, and frequently hold their wings up or down, a sign of flight muscle defects (Fig. 2A,B). The larval fat body fails to turn over within the first 2 days of adult life, another indication of a metabolic defect (Fig. 2C). Mutant flies bearing the weaker clu^d00713 or clu^EP0969 alleles are viable but the males are sterile. Flies bearing combinations of clu alleles, or trans-heterozygotes with a deficiency for the clu region, failed to complement, indicating that these defects are the result of disruption of clu expression. Finally, the Clu protein trap line CA06604 (Buszczak et al., 2007) produces a Clu-fusion protein of the expected size (Fig. 1C) that is likely to be functional because CA06604 homozygotes are viable and fertile.

Flies that are mutant for park, pink1 or rho7, which encode mitochondrial proteins, exhibit many similar defects (Greene et al., 2003; Wang et al., 2006; McQuibban et al., 2006). To determine whether changes in mitochondrial function underlie the abnormalities in clu mutant flies, we carried out electron microscopy on the flight muscles of 3-day-old clu^d08713 adults. Nearly all the flight muscle mitochondria were swollen and showed severe abnormalities, especially vesiculation/loss of cristae (Fig. 2D,E). Mitochondria in the ovary showed similar but less severe effects, including reduced cristae and swelling (Fig. 2F,G). Mature sperm from sterile clu males were completely immobile (data not shown). In addition, the mitochondrial derivative in elongated spermatids was irregular and swollen (Fig. 2I) when compared with wild-type flies (Fig. 2H). We conclude that clu is required for normal mitochondrial structure and function.

To examine whether Drosophila clu is also needed for mitochondrial positioning, we analyzed mitochondria during oogenesis. Drosophila ovarian follicles arise from stem cells in the germarium of each ovariole and subsequently develop in sequence (Fig. 3A). Changes in mitochondrial localization were observed in clu mutant germ cells beginning as early as the germline stem cells (GSCs). Normally, mitochondria in GSCs are enriched anteriorly, but are also found throughout the cytoplasm (Fig. 3B). In clu^d08713 mutant GSCs, nearly all mitochondria occupy the anterior region of the cell (Fig. 3C). This increased clustering raises the question of whether, during GSC division, the daughter cystoblast and newly formed GSC inherit equal amounts of mitochondria. At mitosis in wild-type GSCs and cystoblasts, mitochondria spread uniformly across the outer spindle and are inherited equally by the daughter cells (Fig. 3D) (Cox and Spradling, 2003). By contrast, mitochondria remained clustered during mitosis in clu^EP0969 early germ cells (Fig. 3E, arrow), suggesting that inheritance may be unequal and that one cell may have to make up for a deficit of mitochondria by increased mitochondrial replication (Cox and Spradling, 2006).

Enhanced clustering was also observed in older germ cells in the absence of normal Clu function. Mitochondria are dispersed in young wild-type germline cysts (Fig. 3F) but were highly clustered in similarly aged cysts that were homozygous for clu^d08713 or clu^EP0969, where they associate with the tips of the fusome (Fig. 3G,H), an organelle that interconnects all the germ cells in a cyst. In growing ovarian follicles, mitochondria normally spread evenly throughout the nurse cell cytoplasm (Fig. 3I), but in similarly aged clu^d08713 follicles nearly all mitochondria cluster in a giant aggregate in each nurse cell (Fig. 3J). clu mutations also caused mitochondrial clustering in some somatic follicle cells in the germarium (Fig. 3K,L).

We next examined the distribution of Clu protein to gain insight into the requirement for Clu in mitochondrial positioning and membrane integrity. Both the N-terminal and C-terminal antibodies, and the Clu-GFP protein trap gave indistinguishable results. Clu protein is found exclusively in the cytoplasm and is often present in discrete bodies or aggregates, as well as throughout the cytoplasm (Fig. 4A,B). Lower labeling is observed in somatic cells. Antibody reactivity is abolished in clu null mutations in all cells (Fig. 4C). Double labeling with anti-Clu and a mitochondrial antibody reveals that multiple mitochondria associate with Clu particles (Fig. 4D); however, Clu protein is only observed outside of mitochondria (Fig. 4E,E’).

Ovarian mitochondria frequently move along microtubules as a result of tethering to both Dhc and Khc motors via linker proteins such as Milt (Cox and Spradling, 2006). Examining ovaries that were stained for Clu and microtubules revealed that Clu particles sometimes associate with microtubules (Fig. 4F). The distribution of microtubules in the ovary did not change appreciably in clu^d08713 nurse cells where mitochondria are highly clustered (Fig. 4G,H).

These observations argue that mitochondrial clustering in clu mutants is not caused by changes in microtubule organization. However, comparing the location of mitochondrial clusters in clu mutants with their location in follicles that are mutant for microtubule motor protein genes or for the milton linker (Cox and Spradling, 2006) strongly suggested that Clu affects the movement of mitochondria along microtubules. In particular, sites of mitochondrial accumulation in clu mutants, such as the proximal side of GSCs (Fig. 3C) and adjacent to the fusome in early cysts (Fig. 3G,H), correspond to sites of microtubule plus ends in these cells (Cox and Spradling, 2006). Consequently, clu mutations cause mitochondria to cluster at the sites of microtubule plus ends in several stages where the polarity of the microtubules that transport mitochondria is known. In other situations, such as in nurse cells, mitochondria are less affected by microtubule motor mutations, and the systems controlling their location, as well as microtubule polarity, have yet to be determined.

Many of the effects ofclu mutation that we observed resembled those described following disruption of park (Greene et al., 2003). Consequently, we examined mitochondria in park mutant follicles to see whether they also form clusters. park¹/Df females are sterile, but follicles develop far enough to reveal strong mitochondrial clustering in the nurse cells of previtellogenic follicles (Fig. 5A,B). As follicles age, mitochondrial clustering worsens (Fig. 5C,D). Examining these clustered mitochondria using light and electron microscopy revealed that some were doughnut shaped (Fig. 5E,F) and some grew larger than normal (Fig. 5F; compare Fig. 2F). park mutants also induced clustering in somatic follicle cells (Fig. 5G,H). During spermatogenesis, only one mitochondrial derivative forms in park¹/Df mutant germ cells (Fig. 5P), as reported previously (Riparbelli and Callaini, 2007). clu^EP0969 forms two derivatives but they appear frayed (Fig. 5N,O). Despite this initial difference, during spermatid elongation the park¹/Df mutant mitochondrial derivative becomes irregular and swollen like those in clu^EP0969 spermatids (Fig. 2I,J).

Because of the similarities between clu and park, we investigated whether these two genes interact genetically. park mutant ovaries still contained normal amounts of Clu protein (Fig. 5I,J), so the clustering observed in these animals is not simply the result of reduced Clu expression. The only difference appeared to be a reduction in the number of Clu particles in the park mutant cells. Females that were heterozygous for either clu or park alone were phenotypically wild type and mitochondria did not cluster in nurse cells from stage 5 follicles (Fig. 5K,L). By contrast, mitochondria were noticeably clustered in all similarly aged follicles from park¹/+; clu^EP0969/+ females (Fig. 5M). However, evidence of severe mitochondrial dysfunction was not observed, and both male and female trans-heterozygotes were fertile. We conclude that clu and park interact genetically to promote mitochondrial clustering.

To further analyze the roles played by Clu and Park we investigated whether clu or park mutant animals contain differences in the major mitochondrial proteins that are encoded in the cell nucleus. Western blots revealed that both Pyruvate dehydrogenase and Complex V-alpha (CVα) chain levels were reduced significantly in null clu^d08713 homozygotes (Fig. 6). This decrease was not observed in clu^EP0969 homozygotes, which retain a small amount of Clu protein. The observed reduction indicates that clu^d08713 mitochondria contain reduced amounts of these major mitochondrial proteins, or that the total number of mitochondria is reduced in clu^d08713 homozygotes.

To try and identify genes that are impacted directly by clu mutation, we used microarrays to analyze stage 10B ovarian follicles bearing the weaker clu^EP0969 allele. Such follicles appear normal and successfully complete oogenesis, however, Clu protein is depleted significantly (Fig. 1C) and mitochondria cluster early in germ cell development. These studies revealed a small number of high-level changes in gene expression between wild-type and clu^EP0969 follicles (Table 2).

Several of the most strongly altered genes protect cells from reactive oxygen species that are generated during mitochondrial metabolism. For example, both CG12896 and GstE1, which are genes that are thought to counteract reactive oxygen damage to thiols and to lipids, respectively (Sawicki et al., 2003), are strongly decreased by clu mutation. Several putative DNA repair proteins, including CG10354, a 5′-3′ exonuclease, CG5181 (SSB1), a single-stranded DNA binding protein, and phr6–4, a (6–4)-photolyase, are all strongly lowered. Apoptosis regulators such as the Rep1 caspase inhibitor (Mukae et al., 2000) and the putative SALL4-homolog CG18446 are reduced as much as tenfold. Several genes involved in mitochondrial metabolism are also changed, including SCAP, which regulates fatty acid production (reviewed by Rawson, 2003). In Drosophila, SCAP and its partner SREBP sense 16-carbon fatty acids and control the transcription of target genes involved in fatty acid biosynthesis in mitochondria. Perhaps in response to altered lipid metabolism, transcripts of several other candidate metabolic genes are also altered, including CG3424 (path), which encodes an amino acid/fatty acid transporter; CG11880, which encodes a putative choline transporter; U26, which encodes a putative aminoadipate-semialdehyde dehydrogenase; and others. Thus, the changes in transcript levels that we observe in relatively mild clu mutant follicles suggest that multiple cellular functions that are linked to mitochondria have been compromised. The results suggest that clu mutation reduces mitochondrial function, enhances oxidative damage, and decreases repair or apoptosis.

Our studies identify a novel Drosophila gene, clu, that is essential for the normal cytoplasmic localization and function of cellular mitochondria. Clu-related proteins are highly conserved, especially within a novel 243-amino acid region (the ‘Clu domain’), which is 85% identical (0 gaps) between Drosophila and human Clu. Clu function also appears to be remarkably conserved, because mutations in orthologs in Dictyostelium, Saccharomyces and Arabidopsis all cause mitochondria to cluster into a restricted cytoplasmic region (Zhu et al., 1997; Fields et al., 1998; Dimmer et al., 2002; Logan et al., 2003). In Drosophila, the Clu pathway is also essential to maintain mitochondrial integrity, whereas Dictyostelium and yeast cells that lack CluA or Clu1, respectively, have no detected functional defects.

Disruptions in a conserved process of mitochondrial maintenance involving the park and pink1 genes underlies many cases of Parkinson’s disease (Abou-Sleiman et al., 2006). We find that Drosophila clu mutations cause phenotypes that strongly resemble those caused by park mutations. Moreover, we showed that mitochondria in park mutants cluster in ovarian cells in a very similar manner to their behavior in clu mutants. In addition, park and clu interact genetically; compound heterozygotes have a mitochondrial clustering phenotype that is not seen in either heterozygote alone. These observations argue that Clu and Park function in the same or closely related pathways, and that the human Clu ortholog KIAA0664 might play a role in Parkinson’s disease. Furthermore, these findings suggest that changes in mitochondrial localization may be of significance in this condition.

Mitochondria are often positioned within cells by motor-dependent movement along microtubules (reviewed by Hollenbeck and Saxton, 2005). Clu protein appears to contact mitochondria, especially in the large Clu particles, many of which are located adjacent to microtubules. Our experiments suggest that clu-induced mitochondrial aggregates are caused by changes in microtubule-based mitochondrial transport. In both the Drosophila brain and ovary, mitochondria associate with transport complexes containing the adaptor protein Milt that are linked to both plus-end-directed motors such as Khc and minus-end-directed motors such as Dhc (Glater et al., 2006; Cox and Spradling, 2006). During oogenesis, loss of Khc, or of one of the Milt isoforms that interacts with Khc, cause mitochondria to cluster in cellular regions that are rich in microtubule minus ends, presumably because countervailing movement toward plus ends has been lost [see fig. 5 in Cox and Spradling (Cox and Spradling, 2006)]. Strikingly, mitochondria in clu mutant ovaries accumulate at predicted sites of microtuble plus ends; these include the proximal region of GSCs (Fig. 3C) and against the fusome in early cysts (Fig. 3G,H). Thus, in some situations, clu mutations may act by interfering with minus-end-directed mitochondrial movement along microtubules. Whether Clu particles correspond to mitochondrial transport complexes that mediate these changes, or whether Clu acts indirectly on mitochondrial positioning could not be determined with certainty from our studies.

In other cells, such as nurse cells, clu mutations also cause pronounced mitochondrial clustering (Fig. 3J), but that clustering is not mimicked by mutations in genes encoding microtubule motor proteins or Milt (Cox and Spradling, 2006). Other systems of mitochondrial localization may predominate in such cells but still be subject to regulation by the Clu pathway. For example, in plant cells, Clu was postulated to control mitochondrial localization by regulating the choice between microtubule-dependent and microfilament-dependent transport (Logan et al., 2003). In particular, the TPR domain of Clu was proposed to repress the interaction of mitochondria with microtubules by competing for binding with the TPR region of Kinesin light chain, thereby allowing transport along actin to predominate. The loss of this postulated interaction through clu mutation predicts that mitochondria would accumulate at microtubule minus ends in these cells. Although the default position where mitochondria accumulate in the absence of Clu might vary depending on the particular cell type and transport systems involved, the use of a Clu-dependent pathway to control mitochondrial subcellular location appears to be widespread.

Our experiments show that mitochondria require Clu to maintain their subcellular location and structural integrity. In the presence of even a relatively mild clu mutation, transcripts from genes involved in mitochondrial function and in protection from oxidative damage are reduced. Consequently, we propose that the normal role of Clu is to function in a pathway that controls the location and activity of mitochondria within the cell. Frequently, mitochondria may need to move to a different subcellular location in order to maximize access to substrates or to mitigate damage from oxidative metabolism. For example, in neurons, microtubule-dependent mitochondrial transport is modulated based on the level of respiratory activity (Miller and Sheetz, 2004). Clu protein would participate in a pathway that senses the internal physiological state of individual mitochondria and transduces this information into homeostatic changes in their positions and metabolic activities. When the Clu pathway is impaired, mitochondria would not move or operate normally, and might consequently suffer damage. This scenario is consistent with the changes that we observed in the inner mitochondrial membranes of Clu mutants, with their reduced levels of mitochondrial enzymes and the observed changes in nuclear gene expression. The greater severity of clu mutations in Drosophila compared with Dictyostelium cluA and yeast clu1 might be because of an intrinsically greater requirement for dynamic mitochondrial positioning in the specialized cells of complex metazoans. This model may provide a rationale for the unexpected effects of microtubule inhibitors on mitochondrial function that were reported recently (Wagner et al., 2008).

Fig. 7 illustrates how the clu and park pathway(s) might link microtubule-based mitochondrial transport, mitochondrial physiology and oxidative damage. Under normal conditions, the clu and park pathway(s) would sense the physiological state of mitochondria and activate appropriate levels of minus-end-directed mitochondrial movement along microtubules. For example, mitochondria that are low in respiratory substrates might move to cellular regions where these substrates are abundant. Mitochondria in need of repair might move close to the nucleus where appropriate repair genes would be induced. In conjunction with active plus-end-directed motors, the result would be that mitochondria move dynamically to locations throughout the cell that are appropriate to their physiological state. However, without functional clu and park pathway(s), or if the local level of toxic metabolic products exceeded a threshold, the mitochondria in question would cease minus-end-directed transport and undergo concerted plus-end-directed movement. In many cells, the major locus of microtubule minus ends is found near the nucleus so that plus-end-directed movement would increase the distance between reactive oxygen production and the nuclear DNA.

During early germ cell development in many organisms, a subset of mitochondria within the germ cell cyst is transported and accumulates in a concerted fashion to form the Balbiani body (Cox and Spradling, 2003; Kloc et al., 2004). It has been postulated that Balbiani body formation is a selective process that rids the germline of mitochondria with defective mitochondrial genomes (Pepling et al., 1999; Cox and Spradling, 2003). In Drosophila, mitochondrial movement leading up to Balbiani body formation takes place by minus-end-directed transport along microtubules (Cox and Spradling, 2003; Cox and Spradling, 2006). Consequently, the Clu pathway might contribute to mitochondrial screening. However, Balbiani bodies of normal size were observed to form in clu mutants, suggesting that Clu function is not utilized, or is redundant, at this crucial stage of germ cell development.

Our studies have several implications for understanding Parkinson’s disease. According to our model, the Clu pathway would contribute strongly to the ability of mitochondria to remain functional during aging, despite the high metabolic requirements and environmental stresses experienced by many tissues. This may be particularly important in neural cells such as those that are compromised in Parkinson’s disease. In a cell whose Clu pathway is compromised, mitochondria would operate less efficiently, suffer more damage and wear out faster because they would spend more time in cellular locations that are not appropriate to their metabolic state. The number of mitochondria producing elevated levels of reactive oxygen species would rise, increasing reactive oxygen damage to the nucleus and other cellular components, thereby leading to greater cell death. Consequently, the level of Clu pathway function within an individual might influence their susceptibility to sporadic Parkinson’s disease and to other late-onset neurological disorders.

Mitochondrial mutations may fall into two classes based on their phenotype and on their relationship to the Clu pathway. Some mutations, such as those affecting mitochondrial ribosomal proteins (Royden et al., 1987) or certain metabolic pathways (Zhang et al., 1999), cause bang sensitivity, but differ phenotypically from clu mutations (Fergestad et al., 2006). By contrast, mutations in park (Greene et al., 2003), pink1 (Wang et al., 2006), rho7 (McQuibban et al., 2006), mitochondrial ATP6 (Celotto et al., 2006) and mitochondrial cytochrome oxidase (Xu et al., 2008) resemble clu mutations in causing general defects in movement, flight, muscle or nerve degeneration, male fertility, and longevity. The effects of this latter class of mutations may be too severe to be compensated for by Clu pathway operation, leading to mitochondrial mis-positioning and increased production of reactive oxygen species. Consistent with this, expression of anti-oxidants partially suppresses the effects of pink1 on neurodegeneration (Wang et al., 2006). Mutations in some of these genes, in addition to park, may also cause mitochondrial clustering and interact genetically with clu. Clearly, further studies of the Clu pathway will deepen our understanding of how mitochondria are maintained in cells and why they sometimes become damaged with age.

All symbols are as described in FlyBase. park¹ (Cha et al., 2005) was obtained from Dr Kyoung Sang Cho, and park^Z472 (Greene et al., 2003) was obtained from Leo Pallanck. Wild-type flies were CantonS, except for those used in Fig. 3C, which were OregonR. All other stocks were obtained from Bloomington or Harvard. All stocks were cultured at 22–25°C on standard food.

Antibodies were raised to amino acids 1–104 and 1301–1448 of Clu using the protocol described in Buszczak and Spradling (Buszczak and Spradling, 2006).

Fattened females and males were dissected and fixed following the protocol in Cox and Spradling (Cox and Spradling, 2003). Primary antibodies were diluted as follows: mouse Complex V, α-subunit (1:500, MitoSciences); mouse 1B1 (1:100, Developmental Studies Hybridoma Bank); and guinea pig Clu-C (1:2000, this work). The following secondary antibodies were used: goat anti-mouse IgG2b and goat anti-mouse Alexa Fluor 488; goat anti-mouse IgG1 Alexa Fluor 568; goat anti-guinea pig Alexa Fluor 568 (1:400, Molecular Probes). For microtubule labeling, fattened ovaries were fixed for 15 minutes in 4% paraformaldehyde in Grace’s medium with 1 mM EGTA. Primary antibodies were a 1:1:1 mix of E7, AA 12.1 and AA 4.3 (1:50, Developmental Studies Hybridoma Bank). For DNA labeling, DAPI was added at 5 μg/ml for 10 minutes before VectoShield (Vector Labs) was added. Confocal analysis was performed using a 63× (NA 1.32) PlanApo lens and Leica TCS SP2 or SP5 confocal microscopes, or a Zeiss LSM510 Pascal confocal microscope.

Adult flight muscles and fattened ovaries were dissected in Grace’s media (Sigma), then processed and imaged as described previously (Cox and Spradling, 2006).

Western blotting was done as described in Wilhelm et al. (Wilhelm et al., 2000) with the following modifications: to isolate protein, adults were ground with a pestle in Laemmli buffer with β-mercaptoethanol (BME), and the sample was boiled for 7 minutes before loading on a 4–15% acrylamide gel. The following primary antibodies were used: goat anti-Clu-N (1:15,000, this work), mouse anti-actin (JLA-20, 1:200, Sigma), mouse anti-Pyruvate dehydrogenase E1 α-subunit (PDF, 1:500, Mitosciences) and mouse anti-Complex V, α-subunit (CVα, 1:10,000, Mitosciences).

RNA was extracted from 100 stage 10B ovarian follicles as described previously (Allen and Spradling, 2008). Microarray analyses were performed in duplicate with RNA samples isolated on separate days, and greater than 90% of the genes that changed significantly were reproducible.

We thank Drs L. Pallanck and Kyoung Sang Cho for materials. Don Fox, Rebecca Frederick, Mary Goll and Vicki Losick made useful comments on the manuscript. Shelley Lloyd and Megan Kutzner provided technical assistance, and Mike Sepanski provided technical assistance with electron microscopy. N. Rusan and M. Peifer provided the protocol for labeling microtubules. A.C.S. is an Investigator of the Howard Hughes Medical Institute. Deposited in PMC for release after 6 months.