Peroxisomes Are Required for Lipid Metabolism and Muscle Function in Drosophila melanogaster

Peroxisomes are ubiquitous organelles that perform lipid and reactive oxygen species metabolism. Defects in peroxisome biogenesis cause peroxisome biogenesis disorders (PBDs). The most severe PBD, Zellweger syndrome, is characterized in part by neuronal dysfunction, craniofacial malformations, and low muscle tone (hypotonia). These devastating diseases lack effective therapies and the development of animal models may reveal new drug targets. We have generated Drosophila mutants with impaired peroxisome biogenesis by disrupting the early peroxin gene pex3, which participates in budding of pre-peroxisomes from the ER and peroxisomal membrane protein localization. pex3 deletion mutants lack detectible peroxisomes and die before or during pupariation. At earlier stages of development, larvae lacking Pex3 display reduced size and impaired lipid metabolism. Selective loss of peroxisomes in muscles impairs muscle function and results in flightless animals. Although, hypotonia in PBD patients is thought to be a secondary effect of neuronal dysfunction, our results suggest that peroxisome loss directly affects muscle physiology, possibly by disrupting energy metabolism. Understanding the role of peroxisomes in Drosophila physiology, specifically in muscle cells may reveal novel aspects of PBD etiology.


Introduction
Peroxisomes are small, spherical, single-membrane bound organelles found in almost all eukaryotic cells. Most biochemical pathways in peroxisomes involve lipid metabolism, including boxidation of fatty acids (FAs), a-oxidation of branched chain FAs, and ether lipid biosynthesis [1]. Peroxisomes also participate in reactive oxygen species (ROS) metabolism. Sources of ROS, such as oxidases that produce H 2 O 2 , and antioxidant enzymes, such as catalase, are present in peroxisomes [2].

Peroxisome Membrane Biogenesis
The formation of a mature, metabolically active peroxisome depends on the activity of Peroxin (Pex) proteins [3]. Peroxisomes can form de novo from the endoplasmic reticulum (ER) or divide autonomously from pre-existing organelles. The membrane for the organelle must be defined and proteins must be inserted into the membrane and the matrix. Pex3 is a peroxisomal membrane protein (PMP) that is required for de novo production from the ER as well as for ongoing membrane protein import from the cytosol [4][5][6][7][8]. During de novo organelle biogenesis, newly synthesized Pex3 localizes to the ER, concentrates into foci, and is carried away from the ER in buds that mature into peroxisomes [4]. Yeast, flies, and human cells lacking Pex3 display a complete block in peroxisome biogenesis [9][10][11]. In addition to the de novo biogenesis pathway, existing peroxisomes can grow and divide [12].

Membrane and Matrix Protein Import
The import of proteins into the membrane and matrix of the peroxisome is a highly orchestrated process involving the action of many Pex proteins [13]. Two models exist for the insertion of peroxisomal membrane proteins (PMPs). In the classical model, Pex19 chaperones newly synthesized, cytoplasmic PMPs to the peroxisome and facilitates their insertion into the peroxisomal membrane [14,15]. An alternative model suggests that all PMPs first insert in the ER membrane and are carried in buds that mature into peroxisomes [16]. These two pathways are not mutually exclusive, and most likely, both are utilized. The majority of Pex proteins are dedicated to the import of proteins into the peroxisomal matrix via two defined pathways. Proteins containing a peroxisome targeting signal (PTS) type 1 (PTS1), consisting of the C-terminal tripeptide SKL or other selected variants [17], are bound by cytoplasmic Pex5, shuttled to the peroxisome, and transported across the membrane [18]. A few proteins contain an N-terminal PTS type 2 (PTS2) motif that is recognized by the Pex7 receptor, and transported into the peroxisome in a similar fashion to PTS1-containing proteins [19]. The PTS2 system is not conserved across species. D. melanogaster and C. elegans appear to have lost the PTS2 system entirely and rely solely on the PTS1 import pathway [20,21]. In addition to its roles in membrane biogenesis and PMP insertion, Pex3 has been implicated in pexophagy and peroxisome inheritance during cytokinesis [22,23].

Peroxisomes and Human Health
Impaired peroxisomal membrane biogenesis or protein import, caused by mutations in various pex genes, lead to a wide spectrum of human diseases, known collectively as peroxisome biogenesis disorders (PBDs) [24]. Peroxisome dysfunction disrupts many biochemical pathways and affects almost every organ system. In Zellweger syndrome (ZS), patients display defects in neuronal development, craniofacial malformations, and low muscle tone (hypotonia). Neuronal dysfunction is clearly present in PBD patients and is thought to be responsible for the observed hypotonia. However, myopathy has also been observed in PBD patients [25,26], suggesting that muscles may be sensitive to peroxisome loss independent of neuronal involvement. There is no cure or effective therapy for ZS and afflicted patients usually die in infancy. The etiology of ZS is still unclear, but accumulation of very long chain FAs (VLCFAs) and low levels of ether lipids have been suggested to play causative roles [24].

Animal Model Systems
Mice with impaired peroxisome biogenesis have been generated and serve as useful animal models for PBDs [27]. These animals show many phenotypes similar to PBD patients, including early death, VLCFA accumulation, neuronal migration defects, and locomotor problems. Drosophila have recently become an attractive system to study peroxisome biology and potentially model PBDs. The majority of human peroxisomal metabolic enzymes have homologs in D. melanogaster [20]. Three recent studies have shown that mutations in multiple Drosophila pex genes produce some phenotypes shared with mouse models and PBD patients [10,28,29].
In order to design effective therapeutics for PBDs, more mechanistic details are needed regarding the effects of peroxisome disruption on whole animal physiology. To this end, we have begun to characterize the effects of disrupting early peroxisome biogenesis on Drosophila physiology. Peroxisome biogenesis can be inhibited by a chromosomal mutation in pex3 or by RNAimediated reduction in pex3 mRNA. In the absence of Pex3, flies exhibit no detectable peroxisomes and show impaired lipid metabolism. Longer acyl chain triacylglycerols (TAGs) are elevated and the level of ceramide-phosphoethanolamine (CerPE), the fly analog of the essential lipid sphingomyelin, is reduced in pex3 mutants. Flies with impaired peroxisome biogenesis specifically in muscle tissues show impaired muscle function, possibly due to altered energy metabolism. Human PEX3 expression fully rescues peroxisome loss in the pex3 deletion mutant. These flies containing ''humanized'' peroxisomes will be a useful model for examining pathological mutations in human PEX3.

Peroxisome Distribution in Drosophila
Drosophila peroxisomes have been visualized previously in various tissues [10,[28][29][30][31][32][33]. We expanded the tissues examined, focusing primarily on larval tissues involved in energy metabolism: the gut, fat body, oenocytes, body wall muscles, and the epidermis. Nutrients are absorbed from the diet as they travels through the gut. Excess energy is stored as triacylglycerol within lipid droplets in the fat body, similar to adipose tissue in mammals. Depending on metabolic demands, lipids are mobilized from the fat body and are likely broken down in oenocytes, hepatocyte-like cells [34].
Larval body wall muscles have intensive energy demands, provided partially through lipid metabolism. Epidermal cells line the larval body wall between the cuticle and muscles.

Characterization of pex3 Mutants
In order to inhibit peroxisome biogenesis, we chose pex3 as our target for mutagenesis. Pex3 is involved in the early steps of de novo biogenesis of peroxisomes from the ER and yeast, flies and human cells lacking pex3 have no detectable peroxisomes [9][10][11]. We generated a pex3 deletion mutant by mobilizing a P-element transposon (P{EPgy2}Pex3 EY19206 ) in the 59 untranslated region of the pex3 gene and screening for imprecise excision events. One imprecise excision allele, pex3 2 , carries a 789 bp deletion containing the first non-coding exon, intron one, and part of exon two including the translational start site and 285 bp of the coding sequence ( Fig. 2A). We used a precise excision of the Pelement, pex3 rev , as well as genomic and cDNA rescue transgenes as controls for phenotypic analysis. pex3 2 mutants reared on standard cornmeal/agar died before the wandering third instar larval stage (Fig. 2D). The larval lethality failed to complement a chromosomal deficiency Df(3L)BSC837, which uncovers the pex3 locus [36]. The pex3 2 lethality also failed to complement the previously reported pex3 1 deletion allele [10]. Genomic and cDNA rescue significantly improved viability of the pex3 2 mutant (Fig. 2D), but did not fully rescue indicating that expression levels or tissue specific expression may not have been fully recapitulated. Additionally, pex3 2 mutant larvae are smaller than age-matched pex3 rev control larvae (Fig. 2E). When grown on grape juice plates, most pex3 2 flies survived to the pupal stage and some escaper adults were observed. These few adult escapers were severely disabled, could not fly, and died within several days.
To determine if pex3 2 makes a protein product, we generated anti-Pex3 antibodies in rabbits using recombinant Pex3-His 6 protein. Anti-Pex3 serum recognizes recombinant Pex3-His 6 and endogenous Pex3 protein in total protein extracts from pex3 rev larvae (Fig. 2B), yet no Pex3 band is visible in extracts from pex3 2 larvae ( Fig. 2B and 2C). The genomic pex3 rescue fragment was sufficient to restore Pex3 protein to detectable levels, similar to wildtype, in pex3 2 (Fig. 2B). Pex3 protein is not detectable by immunoblotting when a UAS-dPex3-Myc transgene was driven by the weak global driver, arm-GAL4 (Fig. 2B). However, the same transgene driven by the strong, global driver, act5c-GAL4, produces levels of Pex3 protein higher than endogenous levels (Fig. 2B). Similarly, act5c-GAL4 driving human pex3 transgene (UAS-hPex3-Myc) also produced Pex3 protein, detectable by anti-Myc antibodies, but at levels lower than Drosophila pex3 (Fig. 2B).

pex3 2 Mutants Lack Peroxisomes
We next examined the effect of Pex3 loss on peroxisome biogenesis. The peroxisomal matrix marker, UAS-eYFP-PTS1, was driven by the weak global arm-GAL4 driver and the hepatocytelike oenocytes cells, were examined in wandering 3 rd instar larvae. Many peroxisomes were visible as punctae in the cytoplasm of oenocytes in pex3 rev wandering 3 rd instar larvae (Fig. 3A), but no punctae are present and eYFP-PTS1 remains diffuse in the cytoplasm in the pex3 2 mutant (Fig. 3B). This result indicates that pex3 2 mutants lack peroxisomes that can import PTS1 cargo. Larvae expressing Drosophila (Fig. 3C) or Human (Fig. 3D) pex3 cDNA in the pex3 2 genetic background recover cytoplasmic punctae, indicating the presence of peroxisomes. These results demonstrate that the impaired biogenesis in pex3 2 is due to the loss of pex3 and that Human Pex3 can replace the function of Drosophila Pex3 and restore peroxisome biogenesis.

pex3 2 Mutants have Altered Lipid Metabolism
The major biochemical pathways in peroxisomes involve lipid metabolism, such as b-oxidation of FAs, ether lipid synthesis, etc. We predicted that lipid metabolism would be altered if peroxisome biogenesis was blocked and tested this prediction by performing lipidomic analysis on pex3 2 mutants using liquid chromatographymass spectrometry (LC-MS). The abundance of major lipid classes, polar lipids, diacylglycerols (DAG), and TAG, were not significantly altered in the pex3 2 mutant compared to the pex3 rev control (Fig. 4A). However, the composition of FAs in storage lipids (TAGs) were changed. Longer acyl chain length TAGs are elevated and shorter acyl chain length TAGs are reduced in pex3 2 (Fig. 4B). Similar changes in TAG chain length were observed in larvae with reduced pex3 (Fig. S1). A UAS-inducible transgene (UAS-pex3.IR) encoding a hairpin RNA that targets pex3 for degradation by RNAi (Fig. 5A) was used to reduce the level of Pex3. When the UAS-pex3.IR was driven by the global GAL4 driver, daughterless (Da G32 ), longer acyl chain length TAGs are elevated and shorter acyl chain TAGs are reduced (Fig. S1). In addition to these changes in TAG acyl chain length, we found that the insect specific lipid CerPE is significantly reduced in pex3 2 (Fig. 4C). Insects use CerPE in place of sphingomyelin, a sphingolipid important in neuronal membranes [37]. This result suggests a previously unknown contribution of peroxisomes to CerPE biosynthesis.
Altered TAG chain length raised the possibility that lipid metabolism might be defective in pex3 2 . This prompted us to examine the ability of the pex3 2 larvae to survive under starvation conditions, where lipid catabolism is required for survival. We found that pex3 2 mutants are hypersensitive to starvation (Fig. 4D) suggesting that peroxisomes, likely through b-oxidation of FAs, contributes to starvation-induced TAG breakdown and overall energy metabolism in the animal.

Global Loss of pex3
Many tissues, including the nervous system, muscles, and liver are affected in PBD patients; however, tissue-specific peroxisomal function has not been significantly examined. We selectively inhibited peroxisome biogenesis in specific tissues by RNA interference to uncover tissue specific roles. Peroxisomal loss, accomplished by pex3 knockdown, was first examined in all tissues for comparison with the pex3 2 mutant. Global expression of UAS-pex3.IR decreased Pex3 levels dependent on the relative strength of the global GAL4 drivers (Fig. 5B). With the weak arm-GAL4 driver, Pex3 levels are slightly reduced from the control (76% of arm.+). The intermediate strength Da G32 -GAL4 driver reduces Pex3 levels to 24% of wildtype (at both 18uC and 23uC). The strong act5c-GAL4 driver further reduces Pex3 levels to 14% of wildtype. Peroxisome abundance, revealed by the peroxisomal matrix marker eYFP-PTS1, is also reduced when UAS-pex3.IR is driven by act5c-GAL4 (Fig. 5E) compared to controls lacking UAS-pex3.IR (Fig. 5D).

Tissue-specific Loss of pex3
We next expressed the pex3 RNAi transgene in a variety of tissues and found that muscles are sensitive to the loss of peroxisomes when screened for viability (Table 1). Peroxisome numbers are severely reduced in larval body wall muscles (Fig. 6B) when UAS-pex3.IR is driven by the Mef2-GAL4 muscle driver compared to controls lacking UAS-pex3.IR (Fig. 6A). Stronger pex3 knockdown in muscles, achieved by co-expression of the RNAi component Dicer (Dcr), resulted in lethality at the pharate adult stage (Fig. 6C). Most of these pharate adults had begun eclosing as indicated by the open operculum, the opening in the pupal case through which the adult emerges. Viability of Mef2.pex3.IR,dcr is thus reduced to 34% 61% compared to 99.6% 60.2% for Mef2. dcr controls (Fig. 6D). Wing expansion is also significantly impaired in Mef2.pex3.IR,dcr flies (Fig. 6G) compared to Mef2.dcr controls (Fig. 6F). Of the Mef2.pex3.IR,dcr flies that successfully eclose, 88% 63% have crumpled wings compared to 0.5% 60.3% for Mef2.dcr controls (Fig. 6H). Finally, Mef2.pex3.IR,dcr decreases locomotion, as determined by a climb test, in which the time required to crawl 5 cm was measured. Mef2.pex3.IR,dcr animals took 11.260.8 seconds to crawl 5 cm while the Mef2.dcr controls needed only 3.360.2 seconds to travel the same distance ( Fig. 6E and Video S1). Knockdown of pex3 with another muscle driver, 24B-GAL4, also results in trapped eclosion, crumpled wings, and decreased locomotion (Table 1 and data not shown). This suggests that the loss of peroxisomes specifically in muscles impairs locomotion and two processes that require muscle function, eclosion from the pupal case and wing expansion. The impaired locomotion observed in flies with reduced pex3 is unlikely to be consequence of inappropriate innervation since evoked excitatory junctional potentials (EJPs) were normal in pex3 2 mutant larvae (Fig. S3). Additionally larval body wall muscle structure and synaptic bouton morphology at neuromuscular junctions appear normal in pex3 2 mutants (Fig. S2).

Discussion
Consistent with previous studies, we have found that peroxisomes are present in all tissues we examined in Drosophila melanogaster (Fig. 1), but the abundance of peroxisomes per cell is not uniform in all tissues [10,[28][29][30][31][32][33]. We have generated a pex3 mutant that lacks detectable peroxisomes (Fig. 3B) and does not survive to adulthood (Fig. 2D), also consistent with a previously reported pex3 allele [10]. Impaired peroxisome biogenesis specifically in muscles impairs muscle function (Fig. 6). Muscle dysfunction is also seen in PBD patients, but is thought to be a secondary consequence of neurological defects. We suggest that muscle function may require peroxisomal metabolism independent from any neurological effects.

Lipid Metabolism
Alterations in lipid metabolism were also observed in pex3 2 mutants (Fig. 4). The primary biochemical pathway in peroxisomes is b-oxidation of FAs. In yeast and plants, all b-oxidation occurs in peroxisomes. In mammals, peroxisomes and mitochondria are metabolically linked and coordinate b-oxidation. FAs with longer acyl chain lengths are preferentially broken down in peroxisomes while medium and shorter chain FAs are preferentially broken down in mitochondria. There is likely a handoff of chain-shortened FAs from peroxisomes to mitochondria. The relative contribution of mitochondria and peroxisomes in Drosophila is unknown. It is likely that longer acyl chain FAs are broken down in peroxisomes since mutants defective in peroxisome biogenesis accumulate small amounts of VLCFAs [10,28]. We also see elevated longer acyl chain TAGs and reduced shorter acyl chain length TAGs in pex3 2 mutants (Fig. 4B). VLCFAs are low abundance lipids and likely below the detection limit of the analysis described here. The lipids measured in our LC-MS analysis represent the major classes present in the larvae. pex3 2 mutants are also hypersensitive to starvation (Fig. 4D). Excess nutrients are stored in the fat body primarily as TAG within storage organelles called lipid droplets. During starvation, lipids are mobilized from the fat body, circulate as DAGs in the hemolymph and are broken down by b-oxidation in target tissues [34]. Although mitochondrial b-oxidation genes are up-regulated during starvation [38], indicating an important role for mitochondrial b-oxidation in the starvation response, our data suggests that peroxisomes are also required for survival during starvation. Starved yeast display a similar requirement for peroxisomes; mutations in peroxisome biogenesis genes decrease yeast survival under starvation conditions [39]. This starvation sensitivity is associated with decreased lipid droplets as pex3D yeast have fewer, smaller lipid droplets [40]. The decrease in lipid droplets could reflect impaired lipid storage and could interfere with lipid mobilization. We did not observe a change in total TAG levels in pex3 2 mutants (Fig. 4A). However, whole larvae were used in our analysis and we cannot exclude the possibility that tissue-specific differences were present. pex3 2 mutants die before the wandering 3 rd instar larval stage when reared on standard cornmeal/agar media. Altering diet, and likely nutrient availability, improves viability, but most flies still die as pupae and do not survive to adulthood. Additional studies using defined media will be required to identify the nutrient(s) responsible for this effect. Beginning with the wandering 3 rd instar stage, larvae are no longer consuming food and must rely on the metabolism of stored nutrients for metamorphosis. It is possible that peroxisomes are required for at least part of this metabolism. For example, there may be specific lipids that are poor substrates for mitochondrial b-oxidation and require peroxisomes for their degradation. In addition, peroxisomes might also be required for specific biosynthetic pathways required for larval and pupal development. For example, plants require peroxisomal b-oxidation to generate active forms of auxin and jasmonates [41,42], raising the possibility that there may be specific D. melanogaster metabolites (e.g. hormones) that require peroxisomes for biosynthesis.
A surprising find in our lipidomics analysis was the reduction in CerPE levels in pex3 2 mutants (Fig, 4C). CerPE is used in insects as an analog of sphingomyelin (SM), which is important for neuronal membrane stability and lateral organization in mammals [37,43].
The larval brain and wing imaginal disk have relatively high CerPE levels [44]. A recent study found that CerPE is critically important in glial cells, specifically in the wrapping of axon bundles [45]. Wrapping glia in Drosophila may play a similar role to myelination in mammalian, e.g. to insulate motor axons. The same group found that the knockdown of genes involved in peroxisome biogenesis (e.g., pex10), peroxisomal b-oxidation (e.g., bifunctional protein), and ether lipid synthesis (e.g., glyceronephosphate O-acyltransferase) in wrapping glia were among the 736 genes that caused locomotor defects similar to knockdown of CerPE biosynthetic genes [45]. The locomotor defects in these mutants may be directly caused by low CerPE levels. Neither CerPE nor SM is directly synthesized in peroxisomes [46][47][48] and the effect of peroxisome dysfunction on CerPE levels is likely indirect and remains to be discovered.

Peroxisomes in Muscles
We have observed that knockdown of pex3 specifically in muscles impairs multiple processes that require muscle function, eclosion (Fig. 6C), wing expansion (Fig. 6E), and climbing (Fig. 6H). These results were obtained using two muscle drivers, Mef2 and 24B-GAL4. GAL4 expression is restricted to muscle cells in the Mef2 driver line [49][50][51]. In the 24B driver line, GAL4 is expressed in somatic muscles and a subpopulation of neurons [35,52]. The expression patterns of these drivers suggest that loss of pex3 in somatic muscles is responsible for the observed phenotypes. However, we cannot rule out contributions from other tissues that express GAL4 in these driver lines.
Our results are consistent with a previous study that showed pex16 mutant adults have impaired locomotion and the degree of impairment worsens with age [10]. Impaired muscle function is seen in PBD patients and PBD animal models, but is thought to be a consequence of neurological defects [24,25,27,[53][54][55]. Our results suggest that peroxisomes are required for muscle function independent of neurological involvement.
In what ways might peroxisomes be required for muscle function? Peroxisomes may be directly involved in FA metabolism for energy production. If so, then longer acyl chain lipids may accumulate in muscle lacking peroxisome, which could lead to toxicity, specifically to mitochondria. Such FA toxicity could  . (B) Immunoblot of total protein extracts from wandering third instar larvae shows that Pex3 protein levels are reduced when UAS-pex3.IR is driven by global GAL4 drivers. The weak, global driver armadillo (arm) driving UAS-pex3.IR does not detectably reduce Pex3 levels. The medium strength global driver, daughterless (Da G32 ), driving UAS-pex3.IR reduces Pex3 levels at 18uC and room temperature. The strong global driver, actin 5c (act5c), driving UAS-pex3.IR reduces the level of Pex3 even further. (C) Adult viability of flies expressing pex3.IR driven by global GAL4 drivers was measured. Weak knockdown (arm.pex3.IR) did not appreciably decrease viability (95.361.0%; n = 190). At 18uC, viability of Da G32 .pex3.IR flies was reduced to 25.261.5% (n = 330). At room temperature, viability of Da G32 .pex3.IR flies fell to 160.5% (n = 444). At both temperatures, some flies died as pupae or pharate adults and remained trapped in the pupal case. Viability of act5c.pex3.IR flies was also reduced (11.966.5%; n = 106). (D) Microscopy of wandering third instar larval epidermal cells shows that peroxisome numbers are severely reduced when UAS-pex3.IR is driven by the act5c GAL4 driver compared to controls lacking the RNAi transgene (E). doi:10.1371/journal.pone.0100213.g005 explain the defects in mitochondrial morphology and activity that have been observed in PBD patients and PBD mouse models [25][26][27]53,[56][57][58][59][60]. Other peroxisomal substrates, such as bile acid intermediates and phytanic acid, can induce mitochondrial damage [61,62]. Alternatively, the absence of peroxisomes may also increase ROS levels leading to mitochondrial damage [60]. Because peroxisomes and mitochondria are metabolically linked, breaking the connection between the organelles may have detrimental consequences for the cell.

Conclusions
Through this study we have gained new insights into peroxisome biology in D. melanogaster and examined the effects of peroxisome loss on animal physiology. Peroxisomes are required for Drosophila development, possibly for their role in lipid breakdown during metamorphosis. Not surprisingly, we find alterations in lipid content when peroxisome biogenesis is impaired. Muscle function is also impaired when peroxisome numbers are reduced specifically in muscle cells. These results raise the possibility that the muscle defects seen in PBD patients may be due to problems within the muscle in addition to the known neurological defects. Exploring peroxisome function in Drosophila may have revealed a previously underappreciated role of muscle in the PBD disease state.

Drosophila Stock Construction
pJM573 and pJM623 were injected into w 2 embryos by Genetivision (Houston, TX). Injected flies were backcrossed twice to w 2 flies and multiple insertions were mapped to the chromosome. pJM875 and pJM877 were injected into VK37 and VK31 embryos by Genetivision (Houston, TX). Injected flies were backcrossed twice to w 2 flies and stocks carrying insertions on chromosome two (VK37) and three (VK31) were established.
Independent P-element excision events using the P{Epgy2}-pex3 EY19206 line were screened for gene deletion events by PCR [67,68]. One imprecise excision allele, pex3 2 , carries a 789 bp deletion (3L: 15, 137, 379.15, 138, 166) covering the first noncoding exon, intron one, and part of exon two including the translational start site and 285 bp of the coding sequence.
Antibody Production and Immunoblots pJM630 was expressed in E. coli BL21(DE3). Soluble Pex3-His 6 was purified by nickel affinity chromatography and used as antigen to produce rabbit polyclonal antibodies (Cocalico). Anti-Pex3 serum was immunodepleted against total protein extract from pex3 2 mutant larvae immobilized on nitrocellulose and affinity purified using recombinant Pex3-His 6 immobilized on nitrocellulose.
Total protein was extracted from ,3 wandering 3 rd instar larvae that were collected fresh or frozen in liquid nitrogen and stored at 280uC. Larvae were homogenized in lysis buffer (25 mM HEPES pH 7.4, 100 mM KCl, 2 mM 2-mercaptoethanol, 1% triton X-100, 2 mM EDTA pH 8.0, Roche complete protease inhibitor). Total protein was quantified by Bradford assay and normalized prior to loading on homemade 4/12% polyacrylamide gels. Gels were transferred to nitrocellulose and probed with anti-Pex3 serum (1:1000 dilution). Goat anti-Rabbit HRPconjugated secondary antibodies (Rockland) were used at a 1:10,000 dilution. Secondary antibodies were visualized by chemiluminescence using LumiGLO (Cell Signaling). Membranes were imaged on a LAS-4000 imager (FujiFilm) and band intensities were quantified in ImageJ.

Imaging
Wandering 3 rd instar larvae were dissected in PBS, fixed in 4% formaldehyde, washed in PBS, and mounted in vectashield mounting media with DAPI (Vector Labs). Images were collected on a Zeiss LSM 510 confocal microscope. GFP was excited with a 488 nm argon laser and a HFT 488 primary dichroic. GFP emission was filtered with NFT 490 and BP 500-550 IR filters before collection. DAPI was excited with a chameleon two photon laser (Coherent) at 720 nm and a HFT KP 650 primary dichroic. DAPI emission was filtered with a 480-520 IR filter before collection.
Embryos were collected on grape juice plates, dechorionated, fixed, and mounted in vectashield with DAPI. Images were collected on a Zeiss Axioplan2 epifluorescent microscope. Images were deconvoluted with MetaMorph (Molecular Devices).
Wings were removed from adult flies in ethanol, mounted in euparal, and incubated overnight at 37uC. Wings, larvae, and pupal cases were imaged on a Leica MSV269 stereoscope.
Lipidomics pex3 2 and pex3 rev flies were reared on cornmeal/molasses media and wandering 3 rd instar larvae were collected for analysis. Glycerophospholipids from homogenized Drosophila larvae of different genotypes were extracted using a modified Bligh and Dyer procedure [71]. Briefly, each sample was homogenized in 800 mL of ice-cold 0.1 N HCl:CH 3 OH(1:1) by vortexing for one minute at 4uC. Suspension was then vortexed with 400 mL of cold CHCl 3 for one minute at 4uC and the extraction proceeded with centrifugation (5 min, 4uC, 18,0006g) to separate the two phases. Lower organic layer was collected and solvent evaporated. The resulting lipid film was dissolved in 100 mL of isopropanol:hexane:100 mM NH 4 COOH(aq) 58:40:2 (mobile phase A). Quantification of glycerophospholipids was achieved by the use of an LC-MS technique employing synthetic odd-carbon diacyl and lysophospholipid standards. Typically, 200 ng of each odd-carbon standard was added per sample. Glycerophospholipids were analyzed on an Applied Biosystems/MDS SCIEX 4000 Q TRAP hybrid triple quadrupole/linear ion trap mass spectrometer (Applied Biosystems, Foster City, CA, USA) and a Shimadzu high pressure liquid chromatography system with a Phenomenex Luna Silica column (26250 mm, 5-mm particle size) using a gradient elution as previously described (Ivanova et al., 2007;Myers et al., 2011). The identification of the individual species, achieved by LC-MS/MS, was based on their chromatographic and mass spectral characteristics. This analysis allows identification of the two FA moieties but does not determine their position on the glycerol backbone (sn-1 versus sn-2). Neutral lipids (DAG and TAG) were extracted by homogenizing weighed larvae samples in the presence of internal standards (300 ng 24:0 DAG and 600 ng 42:0 TAG) in 2 mL 1X PBS and extracting with 2 mL ethyl acetate:trimethylpentane (25:75). After drying the extracts, the lipid film was dissolved in 1 mL hexane:isopropanol (4:1) and passed through a bed of Silica gel 60 Å to remove remaining polar phospholipids. Solvent from the collected fractions was evaporated and lipid film was redissolved in 90 mL 9:1 CH 3 OH:CHCl 3 , containing 10 mL of 100 mM CH 3 COONa for MS analysis essentially as described (Lord et al., 2012). Samples were analyzed in triplicates and lipids are presented as pmol/mg weight for neutral lipids and pmol/mg protein for glycerophospholipids.

Larval Starvation
Approximately 50 mating pairs were placed on 6 cm grape juice (20% grape juice, 2.4% agar, 2% ethanol, 1% acetic acid) plates at 25uC in the dark. Embryos were collected for 2 hours and incubated at 25uC. 6661 hr old larvae were placed in 6 cm petri dishes without vents (Nunc, #150326) on a 3 cm square piece of Whatman blotting paper (GB004) soaked with PBS. Plates were wrapped with parafilm and incubated at 25uC in a humid chamber. Surviving larvae were counted every 24 hours.

Climb Test
Adult flies were placed in 136100 mm test tubes and gently vortexed. Video was captured with a Canon SD750 digital camera until the flies reached the top of the vial. The time required to crawl 5 cm was obtained from each video using VirtualDub (http://www.virtualdub.org). Figure S1 pex3 knockdown causes aberrant lipid metabolism. Total lipid extracts from wandering 3 rd instar larvae were analyzed by mass spectrometry at the Kansas lipidomics research center. (A) MS analysis of larval lipids shows that the levels of polar lipids are unchanged in larvae with reduced Pex3 (Da G32 .pex3.IR,dcr) compared to controls (Da G32 .lacZ). Diacylglycerol (DAG) levels are reduced and triacylglycerol (TAG) levels are elevated in larvae with reduced Pex3 compared to controls. (B) Larvae with reduced Pex3 also have elevated longer acyl chain length and reduced shorter acyl chain length TAG species. (TIF) Figure S2 Neuronal and muscle morphology are unaltered in pex3 mutants. Larvae were dissected, fixed, and stained with Alexa 594-conjugated anti-HRP antibodies. Boutons at muscles 6 and 7 were visualized in pex3 rev control (A) and pex3 2 mutant larvae (B) by fluorescence confocal microscopy. Larvae were also dissected, fixed, and stained with Alex 647-conjugated Phalloidin. One hemisegment was visualized in pex3 rev control (C) and pex3 2 mutant larvae (C) by fluorescence confocal microscopy.

Supporting Information
In C and D, anterior is to the left and the ventral midline is to the bottom. The scale bar for A and B is 10 mm and the scale bar for C and D is 50 mm. (TIF) Figure S3 Excitatory junction potentials (EJPs) at the neuromuscular junction are unaltered in pex3 mutants. Larvae were dissected in HL3.1 [72] with 0.8 mM CaCl 2 . Peripheral nerves were cut and stimulated with a suction electrode. Muscle recordings were taken from muscle 6 in abdominal segments A3, A4, or A5. (A) Representative traces for pex3 rev control and pex3 2 mutant larvae show very similar EJP amplitudes. (B) Average EJP amplitudes 6 standard error of the mean of pex3 rev control (39.864.3 mV) and pex3 2 mutant larvae (35.461.9 mV) were not significantly different (P = 0.38, Student's t-Test).