A model of Costeff Syndrome reveals metabolic and protective functions of mitochondrial OPA3

Costeff Syndrome (OMIM 258501), also called Type III 3-Methylglutaconic aciduria (Type III MGA) or Optic Atrophy 3 disorder, is caused by mutations in the OPA3 gene (Anikster et al., 2001). In addition to MGA, individuals with Costeff Syndrome display optic atrophy and extrapyramidal movement disorders and more variably develop ataxia, spasticity, hyperreflexia, cognitive deficits and/or a loss of visual acuity (Costeff et al., 1989).

Costeff Syndrome is a rare autosomal recessive disorder with fewer than 50 documented cases worldwide. OPA3 is a 179 amino acid protein with mitochondrial leader and targeting signals at the N terminus, but no other recognizable structural motifs (Fig. 1A). Three Costeff Syndrome-causing OPA3 mutations have been identified: 40 patients of Iraqi-Jewish origin homozygous for a splice junction mutation (Fig. 1A; IVS-1); one patient of Kurdish-Turkish origin homozygous for an in-frame deletion of six conserved amino acids (Fig. 1A; Del.); and one patient of Indian origin homozygous for a nonsense mutation (Fig. 1A; Q139X) (Anikster et al., 2001; Ho et al., 2008; Kleta et al., 2002). In addition to OPA3 mutations causing Costeff Syndrome, two OPA3 missense mutations, G93S and Q105E, in heterozygous state have been associated with a distinct and somewhat milder dominant disorder termed Autosomal Dominant Optic Atrophy and Cataract (ADOAC; OMIM 165300) in 14 patients in France (Ferre et al., 2009; Reynier et al., 2004).

Costeff Syndrome is one of five MGA syndromes that share a common signature of markedly increased urinary levels of the organic acid 3-methylglutaconic acid (MGC) (Gibson et al., 1993; Gunay-Aygun, 2005). MGC exists in equilibrium with 3-methylglutaric acid (MGR), which also is increased in Costeff Syndrome patients. As a result, MGC measurements can serve as a proxy for both MGC and MGR. Although the best documented origin of MGC in human metabolism is from mitochondrial 3-methylglutaconyl-CoA, Edmond and Popjak (Edmond and Popjak, 1974) and Schroepfer (Schroepfer, 1981) proposed and/or showed evidence for several extra-mitochondrial sources of MGC, all of which derive ultimately from 3-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA; Fig. 1B) (Landau and Brunengraber, 1985; Poston, 1976).

Although the cellular function of OPA3 is unknown, several lines of evidence suggest a role in the inner mitochondrial membrane (IMM), where the electron transport chain complexes and multiple metabolite and protein import systems are located (Fig. 1B). First, human OPA3 has functional mitochondrial leader and targeting signals, and proteomic analysis has found murine Opa3 to be enriched in the IMM (Da Cruz et al., 2003; Huizing et al., 2010). Second, several other optic atrophy disorders and one form of MGA (Barth Syndrome/MGA II, OMIM 302060) are caused by mutations in proteins required for IMM functional integrity (Huizing et al., 2005). Third, a mouse model of Costeff Syndrome has mitochondrial defects (Davies et al., 2008). Based on this evidence, the current clinical recommendation is for individuals with Costeff Syndrome to avoid medications known to impair mitochondrial function or to increase oxidative stress (Gunay-Aygun et al., 2010).

OPA3 is highly conserved across many species, spanning from fungi to human. In this paper, we describe the embryonic expression of zebrafish opa3 and explore the function of OPA3 by generating opa3 mutants and challenging them in several ways. The zebrafish mutants recapitulate key Costeff Syndrome symptoms, including MGC aciduria, and we show that this excess MGC probably derives from extra-mitochondrial HMG-CoA. We further show that the reduction of MGC levels by Opa3 is dependent on its delivery to mitochondria. Finally, we show that opa3 mutant embryos are sensitive to electron transport chain-inhibiting compounds.

Embryos were from natural crosses and staged conventionally (Kimmel et al., 1995). Except where otherwise indicated, opa3^ZM/ZM and opa3^A76T/A76T embryos and larvae came from respective homozygous parents and wild-type (opa3^+/+) controls were from the same genetic background. All microinjections were performed in one- to two-cell stage embryos. The antisense morpholinos (MO) sequences were as follows: standard MO, CCTCTTACCTCAGTTACAATTTATA; Auh MO, TCTACACCTCACTATAGCCGCCATG; Auh mismatch MO, TCaAgACCTgACTATAcCCcCCATG (the lower cases indicate the five mismatched bases). 5 ng MOs were injected. mRNAs were synthesized using the mMessage mMachine SP6 Kit (Ambion). Whole-mount in situ hybridization (WISH) was standard (Thisse and Thisse, 1998). For WISH on embryos older than 24 hours, N-phenylthiourea (Sigma) was used to suppress pigment development.

WISH-stained 3-day-old embryos were embedded in JB-resin and sectioned at 7 μm. WISH-stained 5-day-old embryos, as well as larvae and adults, were embedded in paraffin and sectioned at 5 μm, then counterstained as indicated (HistoServ; Germantown, MD).

MLTS-GFP fusion constructs were created by inserting the mitochondrial leader and targeting signals of opa3 into pEGFP-N1 plasmids (Clontech), and transfected into human fibroblasts then cultured, counterstained with MitoTracker Deep Red 633 (Molecular Probes) and imaged as previously described (Huizing et al., 2010). Control transfections used the pmaxGFP plasmid. The same CMV promoter was used in both constructs.

One-hundred zebrafish embryos per data point were homogenized in 1 ml PBS, and individual adult zebrafish were homogenized in 40 ml PBS. Organic acids were extracted from 1 ml aliquots using ethyl acetate. MGC was measured by isotope-dilution gas chromatography-mass spectroscopy (GC-MS) as previously described (Kelley and Kratz, 1995). Protein concentrations were determined with a BCA protein assay kit (Pierce). Each panel represents parallel runs, except for Fig. 7C where the Ctrl MO versus Auh MO and opa3^+/+ versus opa3^ZM/ZM were from different runs. Differences in measured MGC levels from different runs probably reflect variations in extraction efficiencies or column conditions.

Analysis of spontaneous forward scoots and turns was carried out as described (Burgess and Granato, 2007). Briefly, high speed video recordings (1000 fps) of 30 s duration were taken of larvae using a DRS Lightning high-speed camera (Del Imaging Systems, Cheshire, CT). The testing arena was lit from above at 650 μW/cm² with a white LED, with an infrared array positioned below the stage providing illumination for recording. Larvae were tested in four groups of 25 at 5, 6 and 7 days. Video recordings were analyzed with the Flote software package. All statistical analysis of larval behavior was performed using SPSS.

For buoyancy measurements, up to 15 adult fish were anesthetized in 240 ml of system water supplemented with 0.025% Tricaine (w/v), transferred to a 250 ml graduated cylinder and their positions (top or bottom) recorded. Fish were promptly allowed to recover in clean system water.

To measure maximum angles of ascent and descent and heights in tanks, adult fish were pre-adjusted to 250 μW/cm² lighting for 1.5 hours, then individually transferred to a 1 l tank, pre-adjusted for 5 minutes then filmed for 5 minutes at five frames per second using a microEye IDS-1545LEM camera (1stVision, Andover, MA) and analysis was done with Image J software. Heights were calculated as the average of all 1500 frames. Maximum angles of ascent and descent represent summed averages from the five frames showing the largest angle of ascent and the five frames showing the largest angle of descent.

300 mg of leucine (Sigma) was dissolved in 5 ml 1 N HCl and the pH was adjusted to 7 with 2 N NaOH. A 12 mg/ml solution of leucine in embryo medium [0.006% red sea salts (w/v) and 0.0001% methylene blue (w/v), pH∼6.4] was injected into embryos (∼1.5 nl per embryo) at the one- to 2-cell stage. Embryos were further incubated in a 2 mg/ml solution of leucine in embryo medium. For embryonic leucine measurements, 50 embryos per data point were collected in 100 μl embryo medium and homogenized. The lysate was deproteinated by mixing with 80 μl Li Citrate buffer (pH 2.2) and 20 μl 33% sulpho-salicylic acid, cleared through a 3KMWCO Viva-spin ultra-filter and measured on a Biochrom 30 amino acid analyzer.

Four 14-somite stage embryos in their chorions were placed in each assayed well of an XF24 islet capture plate in 500 μl filtered (0.22 μm) medium (0.006% sea salt), covered with rings (mesh side up) and loaded into a Seahorse XF24 analyzer set at 30°C. The rate of oxygen flux was measured before and after the addition of 75 μl each of the following compounds: oligomycin (700 μg/ml), FCCP (15 μM) and antimycin A (60 μg/ml).

Fifteen 30-hour-old embryos per data point, either untreated or antimycin A treated, were harvested and extracted for ATP as previously described (Yang et al., 2002). ATP concentrations were immediately measured using the Enliten ATP Assay Kit (Promega). The ATP production in untreated opa3^+/+ controls was defined as 100% and used to normalize the ATP concentrations in other samples.

For measuring drug sensitivity, fifty four-cell stage embryos were incubated in a 100 mm×20 mm glass dish containing 30 ml embryo medium, with or without supplementation of 20 nM rotenone, 1 ng/ml antimycin A, 30 μM potassium cyanide (KCN), 0.2 μg/ml oligomycin, 80 nM staurosporine or 0.5 μM simvastatin (all from Sigma). Phenotypes were scored and photographed at 1 day. For opa3^ZM/+ incrosses, individual embryos were genotyped after scoring. For MGC measurements after simvastatin treatment, 100 embryos per dish were treated as above and collected at 1 day for MGC measurements.

Human OPA3 is expressed ubiquitously in adults, but symptoms of Costeff Syndrome suggest particular requirements in the nervous system (Anikster et al., 2001). To determine when and where Opa3 might act during zebrafish development, we localized opa3 transcripts by whole-mount in situ hybridization (WISH). opa3 is expressed in 4-cell stage embryos (Fig. 2A), prior to the onset of zygotic transcription, indicating that the mRNA is present in oocytes. Transcripts of opa3 are ubiquitously distributed through the tailbud stage of development (Fig. 2B,C) then become progressively enriched in the brain (Fig. 2D). This enrichment persists through 5 days of development, when enriched opa3 expression also becomes apparent in internal organs (Fig. 2E,F). Sections of opa3-stained 3-day-old embryos reveal concentrations in the tectum, preoptic region, optic tract, optic nerve and other regions of the retina, namely retinal ganglion cells, inner and outer plexiform layers and retinal pigment epithelium (Fig. 2G-H). Higher magnification views of opa3-labeled optic nerves show that opa3 mRNA is localized in the axons of the optic nerve (Fig. 2I). Sections of opa3-stained 5-day-old embryos also reveal transcripts in the inner ear, swim bladder, heart, liver and intestinal epithelium (Fig. 2J-K).

We recently showed that human OPA3-GFP fusion proteins are targeted to mitochondria in human fibroblast cells (Huizing et al. 2010). There are four amino acid differences between the human OPA3 mitochondrial leader and targeting sequences (MLTS), and the analogous portion of zebrafish Opa3 (Fig. 1A). The putative zebrafish MLTS is not recognized as a mitochondrial signal by prediction software (http://psort.nibb.ac.jp/); however, we find that it successfully targets GFP to human fibroblast cell mitochondria (Fig. 3A-F). We also observe that microinjection of 200 pg opa3 mRNA into zebrafish embryos causes developmental arrest and delays in an MLTS-dependent manner (Fig. 3G-J). Thus, the MLTS of zebrafish Opa3 is sufficient for protein localization to mitochondria and is necessary for the gain-of-function effects of Opa3 in embryos.

To determine the developmental requirements for Opa3, we obtained two mutant opa3 alleles. The ZM_00365842 (opa3^ZM) allele has a 5.2 kb retroviral DNA insertion just downstream of mitochondrial leader signal of opa3 (Fig. 1A). This allele was identified by sequencing insertion sites in a library of retrovirally mutagenized sperm (Jao et al., 2008). The line was then derived by in vitro fertilization (performed by Znomics) and we verified the disruption by sequencing the viral-host DNA junctions (data not shown) and by RT-PCR analysis (see Fig. S1 in the supplementary material). The predicted opa3^ZM product comprises the first 19 amino acids, followed by five virally encoded amino acids and a virally encoded termination codon. Our RT-PCR analysis indicates that erroneous splicing of the opa3^ZM allele causes inclusion of at least 60 nucleotides from intron 1, which introduces additional termination codons in all three frames (see Fig. S1 in the supplementary material). A second mutant allele (opa3^A76T) was derived by TILLING and carries a guanosine 226 to adenosine substitution, converting a conserved alanine (residue 76) to threonine (Fig. 1A; see Fig. S1A in the supplementary material) (Sood et al., 2006).

To determine whether Opa3 deficiency leads to increased levels of MGC in zebrafish, we compared MGC levels in opa3^+/+, opa3^ZM/ZM and opa3^A76T/A76T embryos using isotope-dilution gas chromatography-mass spectroscopy (GC-MS). A substantial increase in the level of MGC was detected in 5-day-old mutant embryos and a more moderate increase was found in mutant adult fish (Fig. 4A). The magnitude of these increases is comparable with that seen in Costeff Syndrome (Elpeleg et al., 1994). No MGC increase was seen in heterozygous 5-day-old embryos (Fig. 4A). Therefore, as in Costeff Syndrome, increased MGC is a recessive phenotype. Although we observed comparable elevations of MGC in opa3^ZM/ZM and opa3^A76T/A76T mutants, we consider the opa3^A76T allele to be a hypomorph, because it fails to produce other phenotypic features we observe in opa3^ZM/ZM mutants. Confirming that the MGC increase is specific to Opa3 deficiency, injection of 25 pg opa3 mRNA suppressed the MGC increase in opa3^ZM/ZM embryos (Fig. 4B).

Our ability to suppress the MGC increase with exogenous opa3 mRNA allowed us to test the functional requirement for Opa3 localization to mitochondria. We found that opa3 mRNA lacking the MLTS (Δmlts) is unable to suppress the MGC increase in opa3^ZM/ZM mutants (Fig. 4B). Therefore, the effect of Opa3 on MGC levels depends on its localization to mitochondria.

Individuals with Costeff Syndrome display optic nerve pallor and retinal thinning from an early age. No overt morphological phenotypes were observed in opa3^ZM/ZM or opa3^A76T/A76T embryos or adults (see Fig. S2A-C in the supplementary material and data not shown), and the optic nerves and other tissues of 6-day-old opa3^ZM/ZM larvae appeared normal in histological sections (see Fig. S2D-E in the supplementary material). However, sections of 1-year-old opa3^ZM/ZM adults revealed a significant reduction in optic nerve diameters (Fig. 5A, white arrows; Fig. 5B). We also observed lesions in 4 out of 6 opa3^ZM/ZM optic nerves (Fig. 5A, arrowheads), with comparable lesions in only one out of 12 optic nerves from other genotypes. We further noted a moderate decrease in the number of retinal ganglion cell bodies, which are the source of the axons comprising the optic nerve (Fig. 5A, black arrows; Fig. 5C). This indicates that the reduced optic nerve diameter is at least partially due to a reduced number of contributing axons.

Although opa3^ZM/ZM mutants are overtly indistinguishable from opa3^+/+ embryos, we noted a small fraction (12 out of 225) with dramatic alterations in swimming behavior. These affected fish position themselves vertically and swim towards the bottom of the tank, rising back towards the top of the tank between efforts (Fig. 6A; see Movie 1 in the supplementary material). This behavior has a variable age of onset, ranging from two to five months. Within 4 weeks of onset, the fish begin to linger close to the water's surface, often upside down. These fish become progressively emaciated and generally perish within 3 months of onset of the behavior.

Based on the intense up-and-down movements of affected fish, we hypothesized that Opa3 has a role in buoyancy control. We compared the neutral buoyancy of opa3^ZM/ZM and control fish after rendering them unconscious with anesthesia. Whereas all control fish sink to the bottom of a test water column, more than 15% of opa3^ZM/ZM fish float to the top (Fig. 6B), including mutants with apparently normal swimming behavior. In a search for more subtle alterations to swimming behavior, we selected adult control (opa3^+/+) and opa3^ZM/ZM mutants with apparently normal swimming behaviors and documented their average positions in individual tanks as well as their maximum angles of ascent and descent. This revealed that maximum angles of ascent and descent are substantially larger among opa3^ZM/ZM mutants (Fig. 6C). No difference was seen in the average tank position.

In light of the grossly impaired swimming in a subset of mutants, we consider the increased angles of ascent and descent across the population of opa3^ZM/ZM fish to reflect a `subclinical' defect in movement or balance control. This may be analogous to the ataxia (uncoordinated movements) seen in a subset of individuals with Costeff Syndrome (Costeff et al., 1989). Although ataxia can be associated with defects in the cerebellum, standard and TUNEL-stained histological sections through an adult opa3^ZM/ZM fish revealed no obvious morphological abnormalities or increase in cell death in the cerebellum or other systems whose disruption might cause movement disorders, namely the inner ear (balance), the hypothalamus (autonomic control of balance and buoyancy), dorsal spinal cord neurons (motor control) or the swim bladder (buoyancy control) (see Fig. S3 in the supplementary material and data not shown).

A characteristic of Costeff Syndrome is the early onset of ophthalmologic and movement defects, so we assessed the vision and movements of zebrafish opa3^ZM/ZM larvae. In addition to ataxia, individuals with Costeff Syndrome varyingly display reduced visual acuity, hyperreflexia, spastic movements and an extrapyramidal movement disorder such as chorea, a form of hyperkinesia, or bradykinesia (Costeff et al., 1989). Zebrafish larvae have a stereotypic set of behaviors with defined kinematic parameters, allowing us to undertake a number of behavioral assays. We identified an early-onset hypokinesia phenotype, reminiscent of the extrapyramidal movement defects seen in Costeff Syndrome. Wild-type zebrafish larvae normally initiate two types of movements: scoots and routine turns (Burgess and Granato, 2007). No significant change in the deployment of routine turns was detected. However opa3^ZM/ZM larvae display a significant reduction in frequency of the initiation of scoot movements (Fig. 6D). Because reduced spontaneous movement initiation is analogous to human hypokinesia, which can result from reduced levels of tyrosine hydroxylase (a dopamine synthetic enzyme), we examined expression of tyrosine hydroxylase in opa3^ZM/ZM embryos, but saw no meaningful difference from controls (see Fig. S4A in the supplementary material) (Wevers et al., 1999). We also asked whether an increase in the level of MGC can itself cause hypokinesia, but saw no changes in the scoot initiation of opa3^+/+ or opa3^ZM/ZM embryos in whom endogenous MGC levels were increased by the addition of exogenous MGC (see Fig. S4B in the supplementary material). The other behavioral assays we ran indicated that opa3^ZM/ZM larvae have normal visual function (see Fig. S2F-H in the supplementary material) and do not suffer from early onset ataxia, hyperreflexia or spasticity (see Fig. S4C-G in the supplementary material).

In summary, our behavioral examination of adult and larval opa3^ZM/ZM mutants revealed a progressive loss of buoyancy control in a subset of fish, as well as more penetrant behaviors analogous to the ataxia and extrapyramidal disorders (bradykinesia) observed in Costeff Syndrome, but not the vision loss, hyperreflexia or spasticity.

Although various mitochondrial and extra-mitochondrial sources for MGC have been proposed and studied over the past three decades, the source of excess MGC in Costeff Syndrome and most other MGA Syndromes remains unclear. To clarify the origin of excess MGC in opa3^ZM/ZM mutants, we performed two biochemical interventions: leucine supplementation and inhibition HMG-CoA reductase, as explained below.

The source of excess MGC is known for one MGA syndrome: Type I MGA (OMIM 250950). In Type I MGA, mitochondrial catabolism of leucine is disrupted by mutations in the AU RNA-binding protein/enoyl-Coenzyme A hydratase, and excess MGC is produced from the consequent build up of mitochondrial 3-methylglutaconyl-CoA (IJlst et al., 2002). Consistent with this, 3-hydroxyisovaleric acid, derived from 3-hydroxyisovaleryl-CoA (Fig. 1B), is also elevated in the urine of individuals with Type I MGA; interestingly, 3-hydroxyisovaleric acid is not increased in Costeff Syndrome (Type III MGA) or other MGAs, suggesting that a leucine-independent mechanism underlies MGC elevation in these diseases (Gibson et al., 1993). To test the hypothesis that Opa3 acts outside of the leucine catabolic pathway, we compared the efficiency of leucine catabolism in opa3^ZM/ZM mutants and a transient model for Type I MGA that we generated using antisense morpholino oligonucleotides (MOs) to block translation of the zebrafish auh ortholog.

Like opa3, maternal auh is expressed in four-cell embryos (see Fig. S5A in the supplementary material), then ubiquitously as embryonic auh until the end of gastrulation (see Fig. S5B-C in the supplementary material). Later, it becomes progressively enriched in the brain and viscera from 1-5 days of development (see Fig. S5D-F in the supplementary material). Histological sections of 3-day-old embryos reveal that auh is expressed in the optic nerve and the same retinal compartments as opa3. Injection of embryos with an Auh MO caused reductions in head and eye size, and elevated MGC levels (see Fig. S5H-J in the supplementary material).

Auh morphants displayed a substantial increase in their leucine levels in response to leucine supplementation (Fig. 7A). By contrast, leucine levels in opa3^ZM/ZM embryos did not change with leucine supplementation. Thus, consistent with normal 3-hydroxyisovaleric acids levels in individuals with Costeff Syndrome, Opa3 is not required for normal leucine flux.

Based on this finding, we turned our focus to potential extra-mitochondrial sources of excess MGC in opa3^ZM/ZM mutants. As described by Schroepfer (Schroepfer, 1981), MGC theoretically could be derived from extra-mitochondrial HMG-CoA by one or more mechanisms, such as the `Popjak shunt' from dimethylallyl-pyrophosphate (PPi) first proposed by Edmond and Popjak (Edmond and Popjak, 1974) (Fig. 1B). This and other proposed pathways would require conversion of extra-mitochondrial HMG-CoA to mevalonate by HMG-CoA reductase (Fig. 1B), and so we predicted that inhibition of HMG-CoA reductase would eliminate the MGC increase in opa3^ZM/ZM mutants. Treatment of zebrafish embryos with the HMG-CoA reductase inhibitor simvastatin caused a shortened body axis, which we could prevent by co-incubation with mevalonate, as reported by others (Thorpe et al., 2004). We saw no phenotypic differences between simvastatin-treated opa3^+/+ embryos and opa3^ZM/ZM mutants (data not shown). To our surprise, we observed no reductions, but rather an increase, in the MGC levels of HMG-CoA reductase-inhibited opa3^ZM/ZM mutants (Fig. 7B). We also observed an MGC increase in simvastatin-treated opa3^+/+ embryos, consistent with reports of MGC increases in humans undergoing statin therapy (Phillips et al., 2002).

The above data argue against excess MGC in opa3^ZM/ZM mutants stemming from the leucine catabolic or mevalonate pathways. However, a third possibility was suggested by our measurements of MGC in leucine-supplemented embryos. When leucine catabolism is unimpaired, as in opa3^+/+ and opa3^ZM/ZM embryos, leucine supplementation is predicted to cause an increase in acetyl-CoA and acetoacetate inside and outside mitochondria, with a corresponding increase in extra-mitochondrial HMG-CoA (Fig. 1B). Leucine supplementation causes an MGC increase in opa3^ZM/ZM, but not opa3^+/+ embryos (Fig. 7C). Together with our simvastatin data, this implies that OPA3 functions to facilitate metabolism of cytosolic MGC derived from extra-mitochondrial HMG-CoA through a novel, mevalonate-independent `salvage' pathway. Unexpectedly, supplemental leucine failed to increase MGC levels in Auh morphants, suggesting the antisense knockdown was incomplete (Fig. 7C).

The electron transport chain (ETC) is a central component of the IMM and consists of five enzyme complexes. Considering the localization of OPA3 to the IMM, it is important to know whether OPA3 influences ETC functionality. The answer is not clear, because two individuals were reported to have ∼50% reductions in ATP levels and a mouse Costeff Syndrome model was shown to have elevated levels of an ETC complex IV component (Cytochrome C Oxidase) in the optic nerve, whereas three patients were reported to have normal ETC complex activities (Anikster et al., 2001; Nissenkorn et al., 1999). We addressed this issue using an extracellular flux analyzer (Wu et al., 2007; Yao et al., 2009). We measured no differences between opa3^ZM/ZM embryos and controls with regard to baseline oxygen consumption (Fig. 8A; media addition), maximal oxygen consumption (Fig. 8A; FCCP addition) or oxygen consumption after addition of ETC inhibitors (Fig. 8A; oligomycin and antimycin A additions). ATP levels were also indistinguishable between opa3^ZM/ZM mutants and controls, both in their basal states and after their reduction by antimycin A treatment (Fig. 8B). Thus, loss of Opa3 does not affect ETC function per se. We also asked whether ETC disruptions affect MGC metabolism, but we observed no changes in the MGC levels of opa3^+/+ or opa3^ZM/ZM embryos after treatment with antimycin A (see Fig. S6A in the supplementary material).

Intriguingly, despite the apparent non-involvement of Opa3 in ETC function, we see a synergistic effect of ETC inhibition and loss of Opa3 function on embryonic morphology. Each of four ETC inhibitors we tested caused variable morphological defects, which we grouped into two classes. Class I embryos have moderate developmental delays (Fig. 8C; cI), whereas class II embryos have more severe delays, small heads and short or irregular tails (Fig. 8C; cII). The ETC inhibitors induced significantly more class II opa3^ZM/ZM embryos than class II opa3^+/+ control embryos (Fig. 8D). As independent confirmation of these data, we applied antimycin A and oligomycin to embryos from heterozygous incrosses of opa3^ZM/+ parents, scored their phenotypes and genotyped them individually, revealing a genotype/phenotype correlation in response to each inhibitor (see Fig. S6B-C in the supplementary material). This combined disruption of Opa3 and the ETC does not affect developmental tissue specification, determined by our observation that eye-, midbrain- and tail-specific markers are robustly expressed in antimycin A-treated opa3^ZM/ZM embryos (see Fig. S6D in the supplementary material).

To test whether the sensitivity of opa3^ZM/ZM mutants to ETC inhibitors is specific, we assessed the effects on opa3^ZM/ZM mutants and controls of two compounds that do not interact with the ETC: the HMG-CoA reductase inhibitor simvastatin and staurosporine, an apoptosis-inducing protein kinase inhibitor. As noted earlier, no differential effects were observed for simvastatin. Staurosporine sensitivity has been reported for fibroblasts from individuals with ADOAC, who carry dominant OPA3 mutations (Reynier et al., 2004). However, we saw no difference between opa3^ZM/ZM and control embryos treated with staurosporine. Thus, opa3^ZM/ZM mutants are specifically sensitive to disruption of the electron transport chain.

Our ability to recapitulate Costeff Syndrome phenotypes by mutation of zebrafish opa3 indicates that the function of this gene has been conserved through much of vertebrate evolution. opa3^ZM/ZM mutants display the three defining features of Costeff Syndrome: MGC aciduria, optic nerve reduction and an extrapyramidal movement disorder (Costeff et al., 1989). The degree of MGC elevation in individuals with Costeff Syndrome is moderate, and we observe a similarly moderate elevation in opa3^ZM/ZM fish. Optic nerve hypoplasia in individuals with Costeff Syndrome is detectable from infancy and often is followed by a loss of visual acuity later in life. By contrast, we detect optic nerve atrophy only in adult opa3^ZM/ZM fish and find no evidence for impaired vision. The mildness of optic phenotypes in opa3^ZM/ZM mutants might be due to the ability of zebrafish, but not humans, to repair optic nerve damage via regeneration (Bernhardt et al., 1996).

The recapitulation of ataxia and an extrapyramidal disorder (hypokinesia in zebrafish; bradykinesia in humans) in zebrafish opa3^ZM/ZM mutants demonstrates conservation for OPA3 at the functional level and is an instructive example of how zebrafish can successfully model specific human movement disorders.

A mouse model of Costeff Syndrome caused by an Opa3 point mutation was recently described (Davies et al., 2008). Opa3^–/– mice display worse optic nerve defects than opa3^ZM/ZM mutants, and a variety of movement disorders, including extrapyramidal dysfunction. The mouse phenotypes are more severe than those in individuals with Costeff Syndrome or in opa3^ZM/ZM mutants, presenting with cardiac myopathy and dying by four months of age, which is less than 25% of a mouse's normal life span. By contrast, individuals with Costeff Syndrome to date enjoy normal life spans (Gunay-Aygun et al., 2010) as do opa3^ZM/ZM mutants, except for the small minority with morbid swimming defects. The overall severity of the mouse model might indicate that fewer redundant or compensatory mechanisms are available to mice than to humans or zebrafish, or perhaps, as one possibility, there is a deleterious gain-of-function component to the murine Opa3 point mutation.

No MGC measurements were made on Opa3^–/– mice because of the difficulty in obtaining urine samples. We were also unable to obtain urine samples from opa3^ZM/ZM mutants, so we measured MGC levels in whole embryos, which are easily generated in abundance. Indeed, the ability to obtain thousands of mutant zebrafish from a few pairs of parents enables running numerous assays on zebrafish models, as we did here. This convenience is bolstered by the ability to generate transient zebrafish disease models with antisense MOs, as shown here for Type I MGA and by others for Type II MGA (Khuchua et al., 2006).

This paper provides several new insights into the functions of OPA3. First, it shows an essential requirement for OPA3 in mitochondria. Although multiple lines of evidence have pointed to a likely requirement for OPA3 in the mitochondrion, here we demonstrate that OPA3 must be delivered to mitochondria to reduce MGC levels.

Second, this paper provides evidence that OPA3 is required to maintain a normal level of MGC in the whole embryo. The precise point of action of OPA3 in MGC biosynthesis has been a long-standing enigma. Here, we confirm a previously elucidated hypothesis that OPA3 acts outside the leucine catabolic pathway, by showing that leucine catabolism is unimpaired in opa3^ZM/ZM mutants (Gibson et al., 1993). Furthermore, we implicate extra-mitochondrial HMG-CoA as the proximal source of the increased level of MGC in opa3^ZM/ZM mutants, suggesting that OPA3 in some manner facilitates the uptake by mitochondria of cytosolic MGC for further metabolism via the leucine catabolic pathway (Fig. 1B). The generation of MGC from cytosolic HMG-CoA might occur via steps of deacylation to free HMG followed by dehydration to MGC. Such an HMG salvage pathway is also suggested by the increased levels of MGC in human patients and in opa3^+/+ embryos treated with simvastatin, and an HMG salvage pathway could also account for increased MGC levels in individuals with Smith-Lemli-Opitz Syndrome, in which a distal block in cholesterol synthesis causes upregulation of HMG-CoA synthase and inhibitory actions of 7-dehydrocholesterol reduce HMG-CoA reductase activity (Kelley and Kratz, 1995; Fitzky et al., 2001; Phillips et al., 2002).

Finally, the distribution of embryonic opa3 mRNA in the retina suggests there are regulatory mechanisms for ensuring the adequate supply of OPA3 to mitochondria. Studies have shown that mitochondrial activity in the human eye is highest in the optic nerve, the retinal ganglion cell layer, the inner and outer plexiform layers, the photoreceptor cell layer and the retinal pigment epithelium (Andrews et al., 1999). Strikingly, we observe opa3 mRNA enrichment in the optic nerve and all but one of the retinal layers (photoreceptors) that are marked by high mitochondrial activity in humans. Based on this apparent colocalization of opa3 mRNA and high mitochondrial activity, we suggest that the transcription and intracellular distribution of opa3 mRNA is regulated to meet mitochondrial demand. Intriguingly, we observed the same pattern of optic nerve and retinal layer distribution for auh mRNA, which also encodes a mitochondrial protein (see Fig. S5 in the supplementary material). It will be of interest to learn whether other opa3 or auh-mRNA-enriched tissues, such as the opa3-enriched intestinal epithelium, are characterized by high mitochondrial activity.

The enrichment of opa3 mRNA in the optic nerve correlates with the occurrence of optic atrophy. Axonal localization of mRNA is uncommon, but has been reported for several other genes, particularly in immature axons (Mohr and Richter, 2000). No other anatomical defects have been reported in individuals with Costeff Syndrome, and we were unable to find any other morphogenetic abnormalities in opa3^ZM/ZM mutants. Nonetheless, the expression of opa3 in the CNS and the inner ear could be related to ataxia, its expression in the inner ear or swim bladder could be related to the buoyancy control difficulties seen in fish, and expression in the heart could be related to the cardiomyopathy in Opa3^–/– mice. No digestive disorders have been observed in individuals with Costeff Syndrome or in animal models, so the medical significance of the expression of opa3 expression in the liver and intestinal epithelium is not clear.

One of the challenges to understanding Costeff Syndrome etiology is to distinguish causes from effects. For example, the abnormal accumulation of some organic molecules has been shown to underlie optic atrophy in several diseases, raising the issue of whether elevated MGC is the cause of optic atrophy or of other neurological disorders in Costeff Syndrome (Huizing et al., 2005). Arguing against this possibility, however, is the observation that individuals with Type I MGA have much higher levels of MGC than do individuals with Costeff Syndrome but do not suffer from optic atrophy. Furthermore, artificially increased levels of MGC failed to induce hypokinesia in wild-type zebrafish larvae. Thus, elevated MGC does not appear to be the cause of optic nerve or behavioral alterations in Costeff Syndrome.

Another question is whether Costeff Syndrome symptoms result from reductions in mitochondrial energy production. This could be true for the retinal ganglion cells that form the optic nerve and certain other cells (see below). However, we see no general abnormalities of energy metabolism of zebrafish embryos that lack Opa3, and we detect no alterations in the MGC levels of zebrafish embryos whose oxidative phosphorylation has been disrupted. We conclude that OPA3 does not have a direct role in oxidative phosphorylation and that the accumulation of MGC is not caused by compromised oxidative phosphorylation. Despite this evidence for non-involvement of OPA3 in oxidative phosphorylation, we observe a potent synergy between loss of OPA3 and ETC inhibition, which supports clinical concerns of mitochondrial sensitivity in individuals with Costeff Syndrome.

Synthesizing data from this study with others, we propose that OPA3 has a role in the mitochondria that is not directly linked to oxidative phosphorylation or MGC metabolism, but rather that OPA3 serves another important role in mitochondrial energy metabolism seated in the IMM. In this model, OPA3 loss-of-function mutations create non-deleterious, hypomorphic biochemical phenotypes in most tissues but overt phenotypes in tissues with a high local demand for mitochondrial ATP production, such as the optic nerve and neurons required for normal movement control; overt phenotypes in OPA3-deficient cells can also arise under other conditions of mitochondrial stress, such as exposure to ETC inhibitory compounds.

The principle that MGC aciduria can arise from an indirect and general mitochondrial defect has been established by Barth Syndrome/Type II MGA, which is caused by a deficiency of Tafazzin, a transacylase that regulates the fatty acyl composition and `maturation' of cardiolipin (Bione et al., 1996). Cardiolipin is a major lipid in mitochondrial membranes, and Tafazzin deficiency leads to cardiolipin depletion, compromising the functional integrity of, principally, the IMM (Bione et al., 1996; Gebert et al., 2009). It is tempting to speculate that OPA3 also serves a structural role in the IMM, enabling optimal function of the electron transport chain and a subset of mitochondrial IMM proteins subserving protein and/or metabolite import.

We thank Forbes D. Porter, Christopher A. Wassif and Lien Ly for assistance with pilot studies; Kevin Bittman, Sung-Kook Hong, Huaibin Cai, Jorge L. Velez and Stephen Wincovitch for technical assistance; Charles P. Venditti and Erich Roessler for helpful discussions. This work was supported by the Intramural Research Program of the National Human Genome Research Institute, the National Institutes of Health. Deposited in PMC for release after 12 months.