A Drosophila model for primary coenzyme Q deficiency and dietary rescue in the developing nervous system

Coenzyme Q (CoQ), also known as ubiquinone, is the only known component of the mitochondrial electron transport chain that is a lipid rather than a protein. It transfers electrons within the inner mitochondrial membrane from Complex I and Complex II to Complex III (for a review, see Hatefi, 1985; Saraste, 1999; Turunen et al., 2004). Together with Complex IV, this generates a proton gradient that is harnessed via the rotary mechanism of ATP synthase (Complex V) to generate ATP (for a review, see Stock et al., 2000; Yoshida et al., 2001). CoQ in its reduced form, ubiquinol (CoQH2), is also a powerful antioxidant preventing reactive oxygen species from damaging mitochondrial and other cellular membranes (for a review, see Navas et al., 2007). Numerous studies have shown that CoQ supplementation can protect mammalian cells from caspase-dependent apoptosis induced by oxidative stress, serum withdrawal and other factors. The anti-apoptotic action of CoQ seems to act upstream of permeabilization of the inner mitochondrial membrane and release of cytochrome c from the intermembrane space (Moon et al., 2005; Papucci et al., 2003; Wu et al., 2007). The ways in which CoQ protects against mitochondrial membrane permeabilization and thus cell death are not clear but might involve scavenging of free radicals as well as a direct interaction with the mitochondrial permeability transition (MPT) pore, a structure that has yet to be defined at the molecular level (Baines, 2009; Fontaine et al., 1998; Lemasters et al., 1998; Papucci et al., 2003; Walter et al., 2002).

CoQ is synthesized from a benzoquinone ring, derived from tyrosine or phenylalanine, which is then linked to an isoprenoid side chain conferring insertion into cell membranes. The most abundant forms of CoQ have side-chain lengths of ten isoprenoid units in humans (CoQ10), nine in rodents and Caenorhabditis elegans (CoQ9), eight in Escherichia coli (CoQ8) and six in Saccharomyces cerevisiae (CoQ6) (for a review, see Turunen et al., 2004). Although human primary CoQ10-deficiency diseases are extremely rare, mutations have been identified in four of the nine genes known from yeast studies to be involved in CoQ biosynthesis (for a review, see Quinzii et al., 2008; Rotig et al., 2007; Tran and Clarke, 2007). Two of these encode prenyl diphosphate synthase subunit 1 (PDSS1) and prenyl diphosphate synthase subunit 2 (PDSS2), subunits of the transferase enzyme responsible for elongating the prenyl side chain of CoQ10. Mutations in the human PDSS1 and PDSS2 genes are both associated with rare infantile multiorgan diseases. For PDSS1, a recessive point mutation in the catalytic domain was linked to early-onset encephaloneuropathy, deafness and livedo reticularis (Mollet et al., 2007; Rotig et al., 2007). For PDSS2, an infant with compound heterozygous mutations presented with nephropathy, myopathy and a progressive neurodegenerative disease called Leigh syndrome (Lopez et al., 2006). Aside from primary CoQ10-deficiency diseases, dietary CoQ10 supplementation has shown beneficial effects in rodent models of neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS), Huntington’s disease and Parkinson’s disease, with small-scale human clinical trials suggesting that it might slow functional deterioration in some cases (reviewed by Galpern and Cudkowicz, 2007; Young et al., 2007).

To develop a genetically amenable model for primary CoQ deficiency, we isolated mutations in a previously uncharacterized Drosophila gene encoding an orthologue of human PDSS1. Using a combination of clonal analysis, caspase-inhibition experiments and dietary CoQ supplementation, we investigated the functions of CoQ and its isoprenoid side chain in the developing larval central nervous system (CNS). These studies provide the first evidence, to our knowledge, that Drosophila can be used to model neural aspects of human CoQ-deficiency disease and their rescue by various CoQ dietary supplements.

Two noncomplementing ethyl methanesulfonate (EMS) mutations, 109 and 264, were isolated in a mosaic screen for genes regulating the size of neural progenitor (neuroblast) clones in the larval Drosophila CNS. Both mutations are embryonic-lethal, and were mapped on the right arm of chromosome 3 by failure to complement molecularly defined deficiencies and a lethal EP element inserted into a previously uncharacterized gene, CG31005 (Fig. 1A; supplementary material Fig. S1A). The predicted protein product of CG31005 (AAF57135) matches that encoded by human PDSS1 (AAH49211) with a BLASTP E-value of 2×10⁻¹⁰². On this basis, and in light of later CoQ rescue studies, we subsequently refer to the CG31005 gene as qless. Sequencing revealed that the qless¹⁰⁹ mutation is a G to A swap at position 1354 of the predicted transcript (CG31005-RA), corresponding to a S215N mutation in the predicted Qless protein (CG31005-PA) (Fig. 1A; supplementary material Fig. S1B). S215 is located ten amino acids C-terminal to the first of the two predicted aspartate-rich DDxxD sites that mediate substrate binding and catalysis by prenyl transferases (Song and Poulter, 1994). Alignments of PDSS1-related prenyl transferases in a wide variety of different organisms shows that position 215 is highly conserved, with either a serine or threonine occupying this position (Fig. 1B). A protein structural model of the region around the first DDxxD motif predicts that the S215N mutation disrupts hydrogen bonding of residue 215 to the N217 side chain, instead resulting in inappropriate hydrogen bonding to Y148 in the N-terminus and to D202 in the catalytic site (Fig. 1C,D). This latter interaction might disrupt catalytic function and, in this regard, it is notable that, like S215, the threonine found at the corresponding position in some PDSS1-related prenyl transferases does not interact with D202 in the structural model (data not shown). Together, these results indicate that S215, located close to the first DDxxD catalytic site, is essential for Qless activity in vivo.

To assess the requirement for PDSS1-like enzyme activity during neurogenesis in the developing Drosophila CNS, we used a FLP-recombinase-based method mosaic analysis with a repressible cell marker (MARCM) (Lee and Luo, 1999). This technique can be used to mark neural clones derived from individual neuroblasts, which divide asymmetrically to self-renew and to generate differentiated neurons and glia (Betschinger and Knoblich, 2004; Sousa-Nunes et al., 2010; Wang and Chia, 2005; Yu et al., 2006). Extensive MARCM analysis of mutant neuroblast clones indicated that the qless¹⁰⁹ and qless²⁶⁴ alleles produce very similar phenotypes. Postembryonic loss of Qless activity in the optic lobe, central brain or thoracic neuromeres resulted in a dramatic reduction in neuroblast clonal growth by 96 hours after larval hatching (ALH) (Fig. 2A–C;,supplementary material Fig. S2). Using Miranda (Mira) as a marker for neural progenitors, a single large Mira⁺ neuroblast was observed in each wild-type clone whereas this cell-type was missing in the majority of thoracic qless mutant clones at 96 hours ALH (Fig. 2D,E; supplementary material Fig. S3). Neuroblast disappearance correlated with a large reduction in postmitotic progeny such that qless mutant clones in the central brain or thoracic neuromeres contained only ∼20% or ∼14%, respectively, of the wild-type number of cells per clone (Fig. 2G). This clearly demonstrates that qless is required for neural growth, but this could be accounted for via primary effects upon cell proliferation, cell survival or both.

Many different protein and nonprotein substrates for prenyl transferases have been described. To determine whether the substrate for the Qless prenyl transferase is indeed CoQ, we induced wild-type and qless mutant clones in larvae raised on a diet supplemented with CoQ10 (see Methods). CoQ10 supplementation did not significantly alter the average number of cells per clone (clone size) in wild-type neuroblast clones at 96 hours ALH (Fig. 2G). This indicates that augmenting the larval diet with CoQ10 cannot boost neural cell proliferation above the wild-type limit. Nevertheless, counts of the numbers of clones per CNS (in central brain or thoracic neuromere regions) revealed a threefold increase in clone frequency (supplementary material Fig. S4). This strongly suggests that CoQ10 supplementation increases FLP recombinase activity and, although the underlying mechanism is not yet clear, it might provide a useful experimental tool for increasing the frequency of FLP-induced mitotic clones. For qless mutant neuroblast clones, dietary supplementation with CoQ10 had a dramatic effect, rescuing the presence of a Mira⁺ neuroblast at 96 hours ALH and the number of cells per clone to near wild-type levels in both the central brain and thoracic neuromeres (Fig. 2F,G). These results demonstrate that qless undergrowth results from a CoQ deficiency and, together with the sequence analysis, they show that qless encodes the predicted PDSS1-like prenyl transferase required for the biosynthesis of active CoQ.

To assess whether the length of the CoQ isoprenoid chain is crucial for rescuing qless activity, we next performed dietary supplementation with CoQ9, the major CoQ isoform in rodents, and also with CoQ4, a synthetic isoform with a side chain that is shorter than that of any of the naturally occurring CoQ major isoforms. As observed for CoQ10, we found that dietary supplementation with CoQ4 or CoQ9 had no significant effect on the size of wild-type neuroblast clones but both CoQ isoforms could efficiently rescue the growth of qless mutant neuroblast clones and the presence of a Mira⁺ neuroblast (Fig. 2G and data not shown). This indicates that CoQ activity in the developing Drosophila CNS can be provided by lipid side chains of four to ten isoprenoid units long.

To investigate the mechanism by which CoQ promotes the growth of neuroblast lineages, we first addressed whether qless is required for cell survival. Immunostaining with CM1 antibody, which recognizes activated caspases in Drosophila, revealed that most neurons present in qless mutant clones at 96 hours ALH were undergoing caspase-mediated apoptosis (Fig. 3A). Cell counting experiments indicated that the percentage of GFP⁺qless cells with activated caspases was 100% in the thoracic neuromeres (n=5 clones) and 98% in the central brain (n=8 clones). Given that CoQ is a known component of mitochondria, we next examined the expression of Hsp60A, a molecular chaperone localizing to the inner mitochondrial membrane, where CoQ participates in the electron transport chain (Baena-Lopez et al., 2008). Hsp60A is known to be upregulated during mitochondrial stress triggered by high temperature or toxin-induced cell death. We found that Hsp60A was strongly upregulated in 100% of the GFP⁺ cells within qless clones (12/12 clones) as compared with surrounding heterozygous GFP⁻ cells (Fig. 3B). Hence, qless is required to protect neural cells against mitochondrial stress and caspase-mediated apoptosis.

In mammals, mitochondrial stresses can trigger release of pro-apoptotic proteins such as cytochrome c from the mitochondrial intermembrane space into the cytoplasm. Although cytochrome c is essential for caspase activation in mammals, it is not required for most types of apoptosis in Drosophila (Dorstyn et al., 2004; Mendes et al., 2006). Drosophila cytochrome c also behaves differently from its mammalian counterpart in that several studies have shown that it remains localized to mitochondria during apoptosis (Dorstyn et al., 2004; Dorstyn et al., 2002; Varkey et al., 1999; Zimmermann et al., 2002). However, this species difference is not absolute as other Drosophila studies show that cytochrome c becomes cytosolic during some types of apoptosis (Abdelwahid et al., 2007; Kanuka et al., 1999). To determine which mechanism occurs during the apoptosis of CoQ-deficient neurons, we examined cytochrome c localization in qless mutant clones. We observed cytochrome c puncta in both wild-type and qless neurons (Fig. 3C). However, punctate cytochrome c expression was markedly increased in qless neurons and this colocalized with the mitochondrial marker Hsp60A (Fig. 3C,D). This suggests that the bulk of cytochrome c remains localized to mitochondria during the apoptosis of qless neurons. As with other aspects of the qless neural phenotype, cytochrome c and Hsp60A upregulation was rescued by dietary supplementation with CoQ10 (data not shown). Together, the mitochondrial analysis demonstrates that Qless prevents mitochondrial stress and promotes cell survival.

To explain fully the qless undergrowth phenotype, it is necessary to distinguish whether CoQ activity is primarily required to prevent cell apoptosis or whether it might also have a direct input into cell growth and cell proliferation. We therefore tested the effects of blocking apoptosis by using MARCM to express p35, an inhibitor of caspase activation, specifically within qless mutant clones (UAS-p35; qless²⁶⁴). Within such clones, we observed some GFP⁺ cells with irregular nuclear morphology and found that punctate cytochrome c remained markedly upregulated in most cells, indicating that p35 does not prevent mitochondrial stress (Fig. 4A). Surprisingly, however, blocking caspase activation was able to rescue efficiently the number of cells per qless mutant clone (Fig. 4B). This implies that the progenitors responsible for generating the rescued cells are themselves also rescued by p35 expression. From this and also our previous observation that qless mutant clones lacked a Mira⁺ neuroblast, we conclude that both the neuroblasts and neurons in these mutant clones undergo caspase-mediated apoptosis. Importantly, rescue of clone size by caspase inhibition also demonstrates that larval activity of qless is required primarily for neural cell survival rather than for cell growth and proliferation.

This study of qless now adds primary CoQ-deficiency diseases to the growing list of available Drosophila disease models. A previous human study showed that the D308 residue within the second DDxxD catalytic site in PDSS1 is necessary for CoQ prenyl transferase activity (Mollet et al., 2007). Our Drosophila Qless analysis has now demonstrated that the serine residue (or threonine in humans) that is located ten amino acids away from the first DDxxD catalytic site is also crucial for CoQ prenyl transferase activity in vivo. Structural modelling suggests that a serine or threonine residue is necessary at this position to prevent inappropriate interactions with the nearby catalytic site. This raises the possibility that PDSS1/Qless activity could be regulated at the ‘first DDxxD plus 10’ position by a serine/threonine kinase. As the human PDSS1 mutation is associated with early-onset neural pathologies, we focused in this study on qless functions in the developing CNS. However, the qless Drosophila phenotype is not restricted to the CNS (data not shown) and so might also prove useful for modelling primary CoQ deficiencies in non-neural tissues. Moreover, as the Drosophila genome also encodes an orthologue of PDSS2, future studies might also shed light on the functions of the other subunit of the mammalian prenyl transferase holoenzyme, which is associated with Leigh syndrome in humans and with nephropathy in both humans and mice (Lopez et al., 2006; Peng et al., 2008).

The primary CoQ deficiency associated with loss of Qless activity induced elevated expression of Hsp60A and cytochrome c, indicating increased mitochondrial stress. This stress was associated with caspase activation and the caspase-dependent apoptosis of neurons. The observation that punctate cytochrome c expression remained elevated when caspase activity was blocked strongly suggests that the sequence of events is the following: loss of CoQ leads to increased mitochondrial stress, in turn triggering caspase activation and thus apoptosis.

We were surprised to observe that blocking caspase activation using p35 was enough to provide efficient rescue of the number of progeny cells generated by each qless mutant neuroblast. This result demonstrates clearly that qless activity is required in a cell-autonomous manner during larval stages primarily for neural cell survival rather than for neuroblast growth and division. The question then arises as to how p35-expressing qless mutant neuroblasts are able to grow and divide in the apparent absence of a normally functioning mitochondrial electron transport chain. One possible explanation is that the ATP needed to power neuroblast growth and division might be mostly generated by CoQ-independent cytoplasmic glycolysis rather than by CoQ-dependent mitochondrial oxidative phosphorylation. Although future neuroblast studies will be needed to test this hypothesis directly, we note that there is a well-studied precedent in the case of cancer cells (Vander Heiden et al., 2009).

The undergrowth of qless mutant neuroblast clones was efficiently rescued by CoQ dietary supplementation. As CoQ supplementation did not lead to enlarged wild-type neuroblast clones, it seems that CoQ in our laboratory larval diet is not growth limiting. Nevertheless, the observation of any undergrowth phenotype in qless clones does indicate that wild-type cells from animals fed on a standard diet do not provide enough CoQ to rescue the growth of their qless neighbours. We found that dietary supplementation with CoQ4, CoQ9 or CoQ10 could all rescue the qless neural undergrowth phenotype. Rescue by CoQ10 is compatible with the finding that CoQ10 can be extracted from insect cell lines (Sagami and Lennarz, 1987), although the CoQ isoform predominating in the Drosophila CNS is not known. A previous study in C. elegans, where CoQ9 predominates, reported that even CoQ6 could partially rescue mutations in the benzoquinone synthetic enzyme Coq7 (Hihi et al., 2003). However, in this case, interpretation of rescue was complicated by possible side-chain elongation of exogenous CoQ6 by resident Qless-like molecules. In the Drosophila qless rescue experiments, such elongation is unlikely and we were therefore very surprised to observe rescue by synthetic CoQ4, with a side chain shorter than all known naturally occurring major isoforms. This suggests that linkage of CoQ to only four isoprenoid units (20 carbons) is sufficient to confer efficient mitochondrial membrane localization in the Drosophila CNS.

In addition to its use in clinical trials for neurodegenerative diseases, CoQ is available as a popular over-the-counter dietary supplement. CoQ levels are reported to decrease with age and it is claimed that supplementation with CoQ10 (often sold as Q10), or variants with shorter lipid side chains, can help to slow dementia, reduce high blood pressure, prevent heart disease and maintain a healthy immune system. Clearly, many more clinical trials are needed to test thoroughly all these bold claims. In the meantime, the Drosophila CoQ rescue paradigm might be useful not only for exploring the basic mechanism of action of CoQ in the developing CNS but also for testing the relative efficacies of the various CoQ derivatives sold as dietary supplements.

Flies were maintained on standard cornmeal/yeast/agar (1% autolysed yeast, 5.8% cornmeal, 5% glucose and 0.6% agar) at 25°C unless otherwise stated. For rescue experiments, standard food was supplemented with CoQ4, CoQ9 or CoQ10 (Sigma-Aldrich) at 50μg per g of food. The embryonic-lethal EMS alleles qless¹⁰⁹ and qless²⁶⁴ were mapped to 100B8;100C1 by failure to complement Df(3R)tll-e and Df(3R)Exel6218. Both qless alleles also failed to complement CG31005^EP984, and the transheterozygote qless¹⁰⁹/qless²⁶⁴ was embryonic lethal (Fig. 1; supplementary material Fig. S1). Most genetic elements used in this study are described in Flybase. For generating MARCM clones, females from the 3R ‘clonemakers’ w, tubP-GAL4, hsFLP¹²², UAS-GFP::6xMyc::NLS; FRT82B, tubP-GAL80^LL3 or w, elav-GAL4^C155, hsFLP¹; UAS-nlsLacZ^20b, UAS-mCD8::GFP^LL5; FRT82B, tubP-GAL80^LL3 were crossed to males of the following genotypes: FRT82B,+/TM6B, Tb, Hu or FRT82B qless¹⁰⁹/TM6B, Tb, Hu or FRT82B qless²⁶⁴/TM6B, Tb, Hu or UAS-p35^BH1/CyO, actin-GFP; FRT82B + or UAS-p35^BH1/CyO, actin-GFP; FRT82B qless¹⁰⁹/TM3,Ser, actin-GFP or UAS-p35^BH1/CyO, actin-GFP; FRT82B qless²⁶⁴/TM3,Ser, actin-GFP. Larvae were raised at 25°C until 47–71 hours after egg laying; clones were then induced at 37°C for 85 minutes in a water bath, and larvae raised at 18°C until the wandering L3 stage, ∼4–5 days later (equivalent stage to 96 hours after larval hatching at 25°C).

The eight exons of CG31005 were amplified by PCR from genomic FRT82B qless¹⁰⁹/FRT + and FRT82B qless²⁶⁴/FRT + DNA in three fragments using an Expand High Fidelity Kit (Roche). DNA sequencing (Cogenics) failed to identify any point mutations in CG31005-PA specific for qless²⁶⁴ but did identify a G to A change at position 1354 in qless¹⁰⁹ (supplementary material Fig. S1). PCR primer pairs: GTATCTTTTCCTTTTCAAGTTTACCAG and CCTGTTGAATTCTGTGCCTGTGTCT (exon 1); AGCATAAAATGTGCGAGGCACTTGG and GGCAAAACTTGTTAGACGTGACAC (exons 2–5); TACGTGTATGGTTTATGTTAGGGTG and GTGTAGGGGAAACATTGCGAATTGC (exons 6–8). Sequencing primers (for exons 1–8): TATCTTTTCCTTTTCAAGTTTACCAG (1), CTTAAGAT TTGCGAT TGGACATGG (2), CGTTAGGTAATAACCACGAGTTC (3), GGGCGTGTGTTTGGCTTTGTTTGCT (4), AGCATAAAATGTGCGAAGGCACTTGG (5), CACACTAATTGCCGAATTAAGCGGC (6), CGATTTTTCTCTGTGCACACAACG (7), TACGTGTATGGTTTATGTTAGGGTG (8).

Tissues were fixed and immunostained essentially as described (Maurange et al., 2008). The primary antibodies used were: 1:50 mouse anti-Mira (gift of Fumio Matsuzaki, Riken Center, Japan), 1:1000 rabbit or mouse anti-GFP (Invitrogen), 1:200 rabbit anti-Hsp60A (gift of Alberto Baena-López, NIMR, UK), 1:200 mouse anti-cytochrome c (clone 7H8.2C12, Invitrogen) and 1:75 mouse anti-CM1 (Invitrogen). Secondary goat antibodies were conjugated to Alexa-Fluor-488 or -594 (Invitrogen) and used at 1:200. Images show projections unless otherwise stated and were collected from a Leica SP5 confocal microscope. MARCM clone cell counts were determined from stacks of confocal sections taken every 1–1.5 μm. As the number of cells per neuroblast clone for qless¹⁰⁹ and qless²⁶⁴ were similar, counts from each genotype were combined for qless-mutant histogram bars. Two-tailed Student’s t tests, assuming equal sample variance, were used to calculate P values.

A search of the sequences in the PDB database (Berman et al., 2000) with CG31005 protein using BLAST (Altschul et al., 1997) revealed the crystal structure of geranylgeranyl pyrophosphate synthetase from Pyrococcus Horikoshii Ot3 (PDB code 1WY0) as a possible template for model building. This was subsequently confirmed by sequence threading using the program GenTHREADER (Jones, 1999). A protein structural model of Drosophila CG31005 was built on geranylgeranyl pyrophosphate synthetase (1WY0) using the modelling software suite QUANTA [Accelrys Software Inc., San Diego, 2006] running on a PC under the Linux operating system. The ‘modeler’ option (Sali and Blundell, 1993) within QUANTA was used to build 20 models of CG31005. The model with the lowest objective function was chosen for further study. The protein structural models for the mouse and human homologues of CG31005 were built in a similar manner (data not shown).

We thank Gary Struhl, Fumio Matsuzaki, Alberto Baena-López and the Bloomington Stock Center for providing fly stocks and antibodies; James Turner for supporting J.G. in the completion of her experiments after joining his laboratory; Louise Cheng, Einat Cinnamon, Rami Makki and Patrick Chinnery for critical reading of the manuscript; and Wai Han Yau for providing figure 1 illustrations. This work was supported by the Medical Research Council (U117584237). Deposited in PMC for release after 6 months.