CMT-associated mutations in glycyl- and tyrosyl-tRNA synthetases exhibit similar pattern of toxicity and share common genetic modifiers in Drosophila

Abstract

Aminoacyl-tRNA synthetases are ubiquitously expressed proteins that charge tRNAs with their cognate amino acids. By ensuring the fidelity of protein synthesis, these enzymes are essential for viability of every cell. Yet, mutations in six tRNA synthetases specifically affect the peripheral nerves and cause Charcot-Marie-Tooth disease (CMT). The CMT-causing mutations in tyrosyl- and glycyl-tRNA synthetases (YARS and GARS, respectively) alter the activity of the proteins in a range of ways (some mutations do not impact charging function, while others abrogate it), making a loss of function in tRNA charging unlikely to be the cause of disease pathology. It is currently unknown which cellular mechanisms are triggered by the mutant enzymes and how this leads to neurodegeneration. Here, by expressing two pathogenic mutations (G240R, P234KY) in Drosophila, we generated a model for GARS-associated neuropathy. We observed compromised viability, and behavioral, electrophysiological and morphological impairment in flies expressing the cytoplasmic isoform of mutant GARS. Their features recapitulated several hallmarks of CMT pathophysiology and were similar to the phenotypes identified in our previously described Drosophila model of YARS-associated neuropathy. Furthermore, CG8316 and CG15599 – genes identified in a retinal degeneration screen to modify mutant YARS, also modified the mutant GARS phenotypes. Our study presents genetic evidence for common mutant-specific interactions between two CMT-associated aminoacyl-tRNA synthetases, lending support for a shared mechanism responsible for the synthetase-induced peripheral neuropathies.

Introduction

Charcot-Marie-Tooth disease (CMT), or hereditary motor and sensory neuropathy, is the most common human inherited neuromuscular disorder affecting specifically the peripheral nerves. CMT symptoms include progressive motor impairment, distal muscle weakness and wasting, sensory loss and skeletal deformities. Based on electrophysiological and neuropathological criteria, the disease is classified into demyelinating forms (CMT1), axonal forms (CMT2), and intermediate forms combining features of CMT1 and CMT2 (Pareyson and Marchesi, 2009). There is no cure for any of these CMT phenotypes, posing substantial personal, social and economic burden. Mutations in over 40 genes are known to cause CMT, making it an extremely genetically heterogeneous disorder (http://www.molgen.vib-ua.be/CMTMutations/). Notably, the known gene products do not represent a single functional category, but range from molecules regulating and establishing the myelin component of the peripheral nervous system (PNS) to ubiquitously expressed proteins. Intriguingly, genetic defects in six aminoacyl-tRNA synthetases (aaRSs) – GARS (Antonellis et al., 2003), YARS (Jordanova et al., 2006), AARS (Latour et al., 2010), KARS (McLaughlin et al., 2010), HARS (Vester et al., 2013) and MARS (Gonzalez et al., 2013) lead to CMT, highlighting the importance of this protein family in PNS homeostasis. Paradoxically, every cell requires the aaRSs function, as they catalyze the aminoacylation of cognate tRNAs during protein biosynthesis, yet mutations specifically affect the maintenance of peripheral neurons.

We reported that mutations in the tyrosyl-tRNA synthetase (YARS) cause dominant intermediate CMT type C (DI-CMTC) (Jordanova et al., 2006) and contributed to the finding that genetic alterations in glycyl-tRNA synthetase (GARS) lead to CMT2D and distal spinal muscular atrophy type V (dSMA-V) (Antonellis et al., 2003). To date, four genetic variants in YARS ( Jordanova et al., 2006; Hyun et al., 2013) and 13 in GARS (referenced in Motley et al., 2010; Lee et al., 2012) have been reported in CMT patients and not in control individuals. In addition to the substitutions observed in humans, two dominant mutations (C201R and P234KY) in Gars cause neuropathy phenotypes in mice (Achilli et al., 2009; Seburn et al., 2006). Both holoenzymes function as homodimers. CMT2D mutations cause a range of changes in the ability of the proteins to homodimerize: some mutations completely abolish dimerization, while others have normal or even enhanced dimerization ability (Nangle et al., 2007). DI-CMTC mutations do not significantly affect the dimerization properties of YARS (Jordanova et al., 2006). Furthermore, the YARS gene encodes for a protein responsible for the charging of tRNA^TYR exclusively in the cytoplasm (the YARS2 paralog is responsible for charging in mitochondria), while the GARS gene encodes both cytoplasmic and mitochondrial forms (Ibba and Soll, 2000). We and others extensively studied the canonical function of YARS and GARS in vitro and in vivo and demonstrated that several of the mutant proteins retain full aminoacylation activity, thereby excluding a loss-of-function mechanism (Nangle et al., 2007; Seburn et al., 2006; Storkebaum et al., 2009). In addition, studies of the Gars^P234KY mouse model have made several putative pathomechanisms considerably less likely. These include haploinsufficiency, dominant-negative effects on enzymatic activity, mistranslation, and protein misfolding with resulting aggregation (Motley et al., 2011; Seburn et al., 2006; Stum et al., 2011). Interestingly, the severity of the phenotype in this mouse is unchanged when crossed to the Wld^S mutation, suggesting that the mechanism of neurodegeneration is distinct from Wallerian degeneration (Stum et al., 2011). Taken together, this suggests that CMT-causing dominant mutations in YARS and GARS (and perhaps other aaRSs) induce a gain of toxic function or interfere with an unknown function of the wild-type (Wt) protein. There are no clues to what this pathologic function could be and whether it is common to both enzymes.

To gain insights into the aaRS-associated disease mechanisms we previously developed a Drosophila DI-CMTC model, representing the first fly model of an inherited peripheral neuropathy (Storkebaum et al., 2009). We observed signs of motor impairment and electrophysiological dysfunction in Drosophila, reproducing the main disease hallmarks. In contrast, mutations in GARS that were identified in humans have not yet been successfully studied at the whole-organismal level in an animal model. Therefore, we generated and characterized flies expressing the G240R mutation, which was identified in a large North-American CMT2D family (Antonellis et al., 2003). This substitution was selected because its pathogenicity is supported by strong genetic linkage data and it has among the most severe CMT2D phenotypes. In order to compare our findings with the well-characterized Gars^P234KY mouse, this variant was also modeled. Importantly, despite targeting neighboring residues, these mutations have different impact on GARS aminoacylation activity: G240R abolishes this function, while P234KY has no effect. We also aimed to compare the toxicity triggered by mutations in GARS and YARS in Drosophila. Finally, we have used our fly models to establish the first genetic link between GARS and YARS in the context of neurodegeneration. While a connection between CMT2D and DI-CMTC has been hypothesized, here we provide experimental evidence that two genes interact in common with CMT-causing variants of both aaRSs.

Materials and Methods

Generation of transgenic flies

The BDGP Drosophila Gene Collection cDNA clone GH09263 was obtained from the Drosophila Genomics Resource Center (Bloomington) and PCR amplified. The forward primer (Aats-Gly-BglIIF) was designed to ligate to the beginning of the cytoplasmic portion of the Aats-gly cDNA (after the mitochondrial targeting sequence) and to add a BglII restriction site and a Drosophila Kozak consensus sequence upstream of the start site; the reverse primer (Aats-Gly-HA-XhoIR) was designed to add a hemagglutinin (HA) epitope tag and a XhoI restriction site. The PCR product was isolated using Qiaquick PCR purification kit (Qiagen) and cloned into the pCR2.1-TOPO (Invitrogen) vector using manufacturers protocols. A wild-type TOPO clone was mutagenized with a protocol adapted from the Quickchange Site Directed Mutagenesis kit (Strategene) using Phusion Taq Polymerase and DpnI (New England Biolabs). Sequencing analysis identified clones with desired mutations. The Wt and mutant constructs were excised from the TOPO vector and purified using the Qiaquick Gel Extraction kit (Qiagen). The inserts were ligated into the pUAST vector (Brand and Perrimon, 1993) using QuickLigase (New England Biolabs). pUAST clones were screened using restriction digestion with HindIII. Clones with insert were isolated using an Endo-Free Maxi Kit (Qiagen) and sequence verified. These clones were then sent to BestGene Inc. for injection and balancing. For each of the three constructs several fly lines were established.

Drosophila genetics and behavioral assays

The UAS-YARS flies used in the modifier screen were described in (Storkebaum et al., 2009). The A307-, ShakB- and c42.2-Gal4 drivers and their expression pattern have been previously reported elsewhere (Allen et al., 1998; Godenschwege et al., 2002; Jacobs et al., 2000). The nsyb-Gal4 driver line (Pauli et al., 2008) was kindly provided by M. Leyssen and B. Dickson (Research Institute of Molecular Pathology, Vienna, Austria). The Act5C-Gal4^strong and Act5C-Gal4^weak driver lines correspond to P{Act5C-Gal4}25FO1 and P{Act5C-Gal4}17bFO1 lines from Bloomington Drosophila Stock Center (BDSC). We obtained from BDSC also the P{EPgy2} and P{XP} enhancer promoter (EP) lines, as well as the following stocks: w¹¹¹⁸;;Gl¹/TM6B, UAS-Apoliner (Bardet et al., 2008), and GMR-htau.Ex (referred as Gl-Tau in the text). The RNAi lines for CG15599 and CG8316 corresponding to lines 108268 and 26956, respectively, were obtained from Vienna Drosophila RNAi Center (VDRC).

All Drosophila crosses were performed at 25°C, 12h light/dark cycle, on standard NutriFly medium (Flystuff).

The insertion sites of P{XP}CG8316^d01774, P{EPgy2}EY06842 and P{EPgy2}EY02909 were determined by inverse PCR following the protocol indicated in the Gene Disruption Project (http://flypush.imgen.bcm.tmc.edu/pscreen/).

Adult flies’ eclosion ratios were determined by the number of Act5C-Gal4>transgene versus the number of Balancer/transgene flies. Lethality phase determination was done according to the description of Bainbridge and Bownes (1981).

Negative geotaxis was performed on adult age-matched female flies, as described in (Storkebaum et al., 2009). The time needed for the fastest fly from the start of the ascent at the vertical wall to a mark at a height of 82 mm was measured. For each group of 10 flies, the experiment was done 10 times and the average of these 10 walking speeds was calculated. For each genotype, 10 groups of 10 flies were tested.

Electrophysiological recordings from the giant fiber circuit

Intracellular recordings from the TTM of adult male and female flies were obtained as previously described (Allen and Godenschwege, 2010). Briefly, the giant fibers (GFs) were activated with 0.03 ms pulses of 30–60 V using two tungsten electrodes inserted into the brain (Grass S44 stimulator, Grass Instruments). Saline-filled glass electrodes were used for recordings from the TTM and a tungsten electrode in the abdomen served as a ground electrode. The recordings were amplified (Getting 5A amplifier, Getting Instruments) and the signals were stored and analyzed using pCLAMP software (Molecular Devices). For the response latencies, individual stimuli were used to determine the time delay between the stimulation of the GFs and the recording of a response in the TTM. The ability of the GF to TTM pathway to follow high frequency stimulations was assessed with 10 trains of 10 pulses given at 100 Hz with a 1 second interval between the trains. The average following frequencies were calculated as percent responses.

Dye injections into the giant fiber

The giant fiber dye-injection and imaging methods have previously been described in detail (Boerner and Godenschwege, 2010; Boerner and Godenschwege, 2011). Briefly, the GF axons of dissected central nervous systems were dye-injected within the cervical connective using a glass electrode (80–100 MΩ) filled with tetramethylrhodamine-labeled dextran (Invitrogen) in 2M potassium acetate by passing depolarizing current. The samples fixed in 4% paraformaldehyde were scanned at a resolution of 1024×1024 pixels, 2.5x zoom, and 0.5 μm step size with a Nikon C1si Fast Spectral Confocal system using a 60x (1.4 NA) oil immersion objective lens. Images were processed using Nikon Elements Advance Research 4.0.

Histological analysis of the giant fiber neurons

The ventral nerve cord of adult female flies was isolated and fixed with 4% PFA in PBS. The A307-Gal4>UAS-Apoliner>UAS-GARS flies were visualized without additional staining. The A307-Gal4>UAS-GARS flies were incubated with monoclonal anti-HA mouse antibody (1:750, Covance). We used Alexa Fluor® 594 goat anti-mouse secondary antibody (1:1000, Life Technologies) to detect the HA tag. The samples were mounted in 90% glycerol supplemented with 0,5% N-propylgallate. Z-stacks were taken on a Zeiss LSM700 confocal microscope (Carl Zeiss) using a LD LCI 25x (0.8 NA) Plan-Apochromat objective with oil immersion and a 40x (1.3 NA) EC Plan-Neofluar oil immersion objective. Image preparation was performed using ImageJ (Schneider et al., 2012).

Retinal degeneration screen

GMR-Gal4; hYARS_E196K/TM6B or GMR-Gal4; GARS_P234KY/TM6B virgins were crossed with EP males. In case of EP male sterility or lethal EP-element insertion, crosses were repeated vice versa, using EP virgins and GMR-Gal4; hYARS_E196K/TM6B males. In F1, female flies heterozygous for GMR-Gal4>hYARS_E196K (or GARS_P234KY)>EP and GMR-Gal4>TM6B>EP genotypes where compared with each other. Positive hits were selected when rough eye phenotype was present in flies with the first genotype and was absent in flies with the second genotype, respectively.

Scanning electron microscopy

Adult flies were terminally anesthetized with ether and directly mounted on aluminum stubs (Electron Microscopy Sciences) without any tissue processing steps. After gold sputter coating, the eyes of the flies were imaged using scanning electron microscopy (SEM) with an SEM505 microscope (Philips).

Western blot analysis

Western blot was performed as described in (Storkebaum et al., 2009) on head protein extracts. We used mouse monoclonal primary antibodies against YARS (1:500, Abnova), HA-polypeptide (1:5000, Covance), α-tubulin (1:10000, Abcam) and horseradish peroxidase-labelled secondary antibodies (Jackson ImmunoResearch Laboratories Inc.). A digital 16-bit image of the chemiluminescent signal was acquired with ImageQuant^™ LAS 4000 (GE Healthcare Life Sciences). Quantification of the band intensities was performed using ImageJ (Schneider et al., 2012). In brief, after background subtraction of the entire image, the lanes were selected and their intensity profiles were plotted with the ImageJ “Analyze > Gels” commands. On these intensity plots of the lanes, the ImageJ magic wand tool was used to measure the area under the peaks as a quantification of the intensity of the band. The intensity of the GARS-HA bands was normalized to the intensity of the α-tubulin bands.

Statistical analysis

The GraphPad Prism 4 and epiR softwares were used for statistical analyses. One-way ANOVA was used to evaluate the statistical significance of the differences in GARS protein expression levels in different transgenic lines. Z-test for one proportion was performed to compare differences in the observed Act5C-Gal4>GARS_Wt vs. Balancer/GARS_Wt ratio, compared to the expected 1:1 ratio. Mantel-Haenzsel analysis was performed to compare eclosion ratios of flies carrying the different mutant genotypes against flies carrying the wild type transgene. For both strong and weak Act5C-Gal4 drivers, odds ratios of Act5C-Gal4>GARS vs. Balancer/GARS per experiment, as well as pooled odds ratios across experiments were calculated to take into account three independent experiments. Cochran’s Q test was used to assess heterogeneity between independent experiments. Unpaired Student’s t-test was used to analyze the negative geotaxis data. One-way ANOVA with Bonferroni’s Multiple Comparison Test was applied to analyze the GF electrophysiology data. The negative geotaxis and GF electrophysiology data are presented as mean ± standard error of mean (s.e.m.).

In silico protein analyses

Expression data and functional annotations were retrieved from FlyBase and modENCODE (Roy et al., 2010). ClustalOmega was used for protein sequence alignment (Sievers et al., 2011). For interspecies orthologue detection we used OrthoDB (Waterhouse et al., 2013), InParanoid (Ostlund et al., 2010), and BLAST (Altschul et al., 1990). Conserved domain prediction was done using InterPro and CDD (Marchler-Bauer et al., 2011). Remote homology search was performed with PSI-BLAST (Altschul et al., 1997), HHPred from the MPI Bioinformatics Tool Kit (Biegert et al., 2006) and FUGUE (Shi et al., 2001). PSI-BLAST was run with default settings, except that low complexity regions were filtered, and using 3 and 4 iterations for CG15599 and CG8316, respectively. Secondary structure and functional predictions were performed with the Jpred3 Web service (Cole et al., 2008) on the alignment of the Drosophila orthologues, and with ELM (Puntervoll et al., 2003). ELM was also used to check for the presence of peptide motifs and transmembrane domains. Subcellular localization was further predicted using ngLOC (King and Guda, 2007), Euk-mPLoc v.2.0 (Chou and Shen, 2010) and CELLO v.2.5 (Yu et al., 2004). We applied FFPred for predicting the function of proteins independent of homology using a feature-based machine-learning approach (Lobley et al., 2008).

Results

Ubiquitous and panneuronal expression of the cytoplasmic form of mutant GARS is toxic in Drosophila

As in mammals, the Drosophila genome contains a single gene encoding both cytoplasmic and mitochondrial isoforms of glycyl-tRNA synthetase (Aats-Gly) (Chihara et al., 2007). The Drosophila Aats-Gly (referred to as GARS in the text) is 60% identical and 75% similar to the human (GlyRS) and mouse (Gars) orthologues. The GARS mutations evaluated in this study affect conserved amino acid residues (Supplemental Fig. 1). We targeted the expression of cytoplasmic forms of Wt and mutant GARS transgenes to different cells and tissues in Drosophila using the Gal4-UAS system (Brand and Perrimon, 1993). Transgenic lines with comparable expression levels were used in the experiments (Supplemental Fig. 2A and 2B).

Initially, we tested the effect of GARS mutations at the whole organism level, by ubiquitously expressing the transgenes with two Act5C-Gal4 drivers (strong and weak). UAS-GARS males where crossed with Act5C-Gal4/Balancer virgins and we counted the number of flies in the F1 generation that eclosed with the two possible genotypes (Act5C-Gal4>transgene and Balancer/transgene). The expected ratio of eclosure for these genotypes is 1:1. Act5C^strong-Gal4>GARS_Wt expression mildly affected the eclosion ratio, while Act5C^weak-Gal4>GARS_Wt ubiquitous overexpression had no effect on viability, and flies eclosed at the expected ratio (Fig. 1A and Supplemental Table 1). Strong GARS_G240R expression severely impaired the fly development at late pupal stage (P15) and yielded only 3% adult escapers, while weak expression had no detrimental effect. Importantly, both strong and weak ubiquitous expression of GARS_P234KY induced lethality in the first instar larval stage and no adult Act5C-Gal4>GARS_P234KY flies emerged (Fig. 1A and Supplemental Table 1). The observed phenotypes were confirmed with 2 independent transgenic lines of each genotype. This data suggests that mutant GARS is toxic when expressed ubiquitously in Drosophila

Fig. 1. Expression of mutant GARS in Drosophila causes pre-adult lethality and induces motor impairment.

A) Ubiquitous expression of GARS_Wt has mild (with Act5C^strong-Gal4) or no effect (with Act5C^weak-Gal4) on the expected 1:1 adult flies’ eclosion ratio. GARS_G240R expression has a dosage-dependent effect on the adult offspring eclosion, while GARS_P234KY expression has detrimental effect upon both strong and weak ubiquitous expression. On the Y-axis the ratio of genotypes eclosing per cross is indicated. Dashed line marks the expected 1:1 genotypes’ eclosion ratio. ***, P<0.001; ns – non-significant were determined after Mantel-Haenzsel analysis which compares the odds ratios (Act5C-Gal4>transgene vs. Balancer/transgene) of flies with ubiquitous expression of wild type and mutant GARS flies of three independent experiments.

B) Panneuronal expression of GARS_G240R with nsyb-Gal4 renders adult flies, which exhibit progressive motor impairment as evaluated in a negative geotaxis assay. The Y-axis indicates the time needed for the fastest fly to climb to a height of 82mm. *, P<0.05 and **, P<0.01, was determined after unpaired Student’s t-test.

We then sought to evaluate putative toxic effects induced by specific expression of GARS mutations in the nervous system of Drosophila, using the nsyb-Gal4 driver. The nsyb-Gal4>GARS_P234KY flies reached late stage of pupal development (P15), but were unable to execute the pupal to adult ecdysis. Nsyb-Gal4-driven expression of GARS_G240R and GARS_Wt yielded viable adult flies, whose motor performance was scored in a negative geotaxis assay. We could not detect significant differences in climbing ability of 10 day-old flies. However, progressive motor impairment of GARS_G240R compared to GARS_Wt animals was observed upon ageing, at day 20 and 30 (Fig. 1B). This suggests that while wild-type GARS is not toxic in the fly nervous system, GARS_G240R and GARS_P234KY are both harmful to neurons. In particular, GARS_G240R is neurotoxic in an age-dependent manner, suggesting that this is a faithful model of CMT2D.

Mutations in GARS affect functionality and morphology of the giant fiber system in Drosophila

To circumvent the observed lethality associated with GARS_P234KY and to test the effect of both mutations, we restricted their expression to the neurons of the giant fiber circuit in adult flies. The appropriately named giant axons of the circuit have already proven useful in modeling CMT in Drosophila, as we demonstrated progressive impairment of synaptic strength and reliability when expressing mutant YARS proteins (Storkebaum et al., 2009). Similarly, we assessed the impact of GARS mutations on GF function upon A307-Gal4 expression in the GFs and their postsynaptic target neurons throughout their development. We recorded the response latency (RL) and the ability to follow 100 Hz repetitive stimulations of the GF-tergotrochanteral motoneuron-tergotrochanteral muscle (GF-TTMn-TTM) circuit (Fig. 2A and 2B), as a measure for synaptic strength and reliability of the GF-TTMn connection (Allen and Godenschwege, 2010). Upon GF stimulation in the brain, we observed a mild but significant RL increase in the GARS_G240R flies (~1 ms) when compared to controls (GARS_Wt and A307-Gal4/+) (~0.8 ms) (Fig. 2C). Unlike the control flies, the GF-TTMn synaptic reliability decreased in one day old GARS_G240R expressing flies (Fig. 2D), and did not progress in 8 and 16 day old flies (data not shown). In contrast, in 40% of the recordings in GARS_P234KY flies, the TTM response was completely absent when the GF was stimulated, suggesting that GF stimulation did not evoke a postsynaptic potential in the TTMn or that the synaptic strength was below the threshold to trigger an action potential. However, in the remaining 60% of the recordings the RL was severely increased to ~1.8 ms (Fig. 2A and 2C), which associated with significantly reduced ability of the GF-TTMn-TTM pathway to follow high frequency one-to-one stimulations already one day after eclosion (Fig. 2B and 2D).

Fig. 2. Electrophysiological and morphological impairment of the GF-TTMn-TTM circuit in mutant GARS expressing flies is not associated with apoptotic cell death or GARS mislocalization and aggregation.

A) Sample traces of recordings from TTM upon GF stimulation (black arrow) in the brain. Control animals have response latency ~0.8 ms (dashed line), while A307-Gal4>GARS_P234KY flies have significantly increased response latency. The response latency of ShakB-Gal4>GARS_P234KY animals is not affected and is similar to controls. Asterisk indicates the recorded response.

B) Sample traces of recordings from TTM upon applying train of 10 stimuli at 100 Hz. Control animals respond (asterisk) one-to-one to each applied stimulus, while A307-Gal4>GARS_P234KY flies have severely impaired ability to follow high frequency stimulations. The ShakB-Gal4> GARS_P234KY animals respond to ~50% of the applied stimuli.

C) Average response latency of wild type and GARS mutant flies upon A307-Gal4 expression. Expression of GARS_G240R and GARS_P234KY in the GF circuit throughout development significantly increases the response latency in comparison to control flies (A307-Gal4/+). In contrast, expression of GARS_Wt has no effect on the response latency.

D) Average following frequency at 100 Hz of wild type and GARS mutant flies upon A307-Gal4 expression. Expression of GARS_G240R and GARS_P234KY significantly decreases the ability to follow stimuli at 100 Hz when compared to control flies or flies expressing GARS_Wt.

A minimum of ten one day old flies (≥20 TTMs) were tested for each genotype in (C) and (D). *, P<0.05; **, P<0.01 and ***, P<0.001, one-way ANOVA with Bonferonni Multiple Comparison Test.

E) Presynaptic (c42.2-Gal4) expression after GF synapse formation or postsynaptic (ShakB-Gal4) expression throughout development of GARS_P234KY flies does not significantly affect the response latency when compared to control flies (Gal4-drivers/+) or flies expressing the transgene with the A307-Gal4 driver. A307-Gal4 data are reused from panel C of this figure.

F) The ability to follow stimuli at 100 Hz significantly decreases when GARS_P234KY is expressed with A307-Gal4, c42.2-Gal4 and ShakB-Gal4 drivers, compared to control flies (Gal4-drivers/+). A minimum of five one day old flies (≥10 TTMs) were tested for each genotype in (E) and (F). *, P<0.05; **, P<0.01 and ***; P<0.001, one-way ANOVA with Bonferonni Multiple Comparison Test. A307-Gal4 data are reused from panel D of this figure.

G) Giant fiber morphology in wild type and mutant GARS flies. Dye injected in the A307-Gal4>GARS_Wt GF axon is homogenously distributed through the axon and synaptic terminal. In contrast to A307-Gal4>GARS_Wt flies, the GF terminal of GARS_P234KY flies is thinner and vacuolar structures (arrows) are observed throughout the axon and synaptic terminal. Scale bar in (G) is 20 μm.

H) TTMn cell of a ten-day old female flies expressing Apoliner (Apo) under the control of A307-Gal4 driver. The eGFP and mRFP fluorescence signals are shown independently and in overlay. Assessment of caspase-induced cleavage of the Apoliner reporter does not reveal apoptosis in the TTMn cell bodies of GARS_Wt and GARS_P234KY flies, as the eGFP signal does not translocate to the TTMn nucleus, but co-localizes with the mRFP signal. eGFP – enhanced green fluorescent protein; mRFP – monomeric red fluorescent protein. Scale bar is 20 μm.

I) GARS-HA immunofluorescence pattern in the giant fiber axons, giant fiber-TTMn synapsing area and TTMn cell bodies upon A307-Gal4 driven expression of GARS_Wt and GARS_P234KY. Both wild type and mutant flies have a comparable distribution pattern of GARS-HA along the GF axon and in the TTMn cell body, without any visible protein aggregation. Scale bar is respectively 20 μm and 10 μm for the GF and TTMn images.

Furthermore, we expressed the GARS_P234KY transgene presynaptically, after the GF to TTMn synaptic contact had been established (c42.2-Gal4), as well as postsynaptically in the TTMn throughout motor neuron development (ShakB-Gal4). Although neither expression affected the RL significantly (Fig. 2A and 2E), the ability of the GF-TTMn to TTM pathway to follow repetitive stimulations was significantly reduced to ~50% when GARS_P234KY was expressed with either Gal4-line (Fig. 2B and 2F). This suggests that GARS_P234KY has a disruptive effect in the GF and the TTMn during development, but also on the synapse function after its development.

In order to determine the GF morphology in GARS_P234KY mutants and GARS_Wt control animals, fluorescent dye was injected into the GF axons. In comparison to GARS_Wt controls, the synaptic terminal of A307-Gal4>GARS_P234KY flies was abnormally thin and the axon as well as the synaptic terminal contained prominent vacuoles (Fig. 2G). These anatomical phenotypes correlated with the physiological defects noted earlier, and were reminiscent to the ones observed in the mutant YARS-expressing animals (Storkebaum et al., 2009).

The electrophysiological and axon-terminal abnormalities were not due to apoptotic cell death, as evaluated by the subcellular localization of the in vivo caspase reporter Apoliner (Bardet et al., 2008) (Fig. 2H). This is in line with our findings that despite the severe functional and morphological synaptic defects, the neurons of the GF circuit were present in all analyzed animals. Consistent with findings in the Gars^P234KY mice, no mislocalization or aggregation of GARS_P234KY was observed and the mutant protein had a comparable distribution with GARS_Wt in the GF and TTMn (Fig. 2I).

GARS and YARS share common genetic modifiers in Drosophila

We have previously shown that retinal expression of YARS_E196K (GMR-Gal4>YARS_E196K) induced a rough eye phenotype in a dose-dependent manner (Storkebaum et al., 2009). Here, we observed that also GMR-targeted expression of GARS_P234KY causes a mild rough eye phenotype (Fig. 3D). Retinal expression of GARS_Wt (Fig. 3E) and GARS_G240R (data not shown) did not affect the ommatidial organization.

Fig. 3. YARS and GARS mutant-specific interactions in the adult fly retina.

Control flies (GMR-Gal4/+) (A) or flies expressing YARS_E196K (B), YARS_Wt (C) and GARS_Wt (E) in the retina do not exhibit rough eye phenotype. GMR-Gal4>GARS_P234KY expression renders mild rough eye phenotype (D). GMR targeted expression of d01774 mildly affects the retina on its own (H), or when co-expressed with Wt YARS and GARS forms (J and L, respectively). In contrast, it provokes a strong rough eye phenotype when co-expressed with YARS_E196K and GARS_P234KY (I and K). Retinal expression of EY06842 per se does not show a phenotype (N), or in Wt YARS and GARS background (P and R, respectively), but gives a strong interaction when co-expressed with mutant YARS and GARS (O and Q, respectively). Retinal expression of RNA_i^CG8316 and RNA_i^CG15599 on their own does not give a phenotype (F and G, respectively). Retinal co-expression of d01774, RNA_i^CG8316 and YARS_E196K fully recovers normal ommatidial organization (M), proving that d01774 is mis-expressing the CG8316 gene. Likewise, simultaneous expression of EY06842, RNA_i^CG15599 and YARS_E196K restores the normal morphology of the fly eye (S). Representative SEM micrographs at lower and higher magnification are presented. Scale bars in (A) are 50 μm.

We made use of the sensitized background induced by one copy of YARS_E196K to perform a gain-of-function screen for genetic enhancers of the retinal degeneration (Rorth, 1996). Flies expressing YARS_E196K transgene in the eye (GMR-Gal4; YARS_E196K/TM6B) were crossed to flies carrying enhancer promoter P-element insertions (EPs), targeting genes on the Drosophila X-chromosome (Bellen et al., 2004; Thibault et al., 2004) and enhancement of the rough eye phenotype was scored in the female animals of F1-progeny, which were heterozygous for the EP-insertion. Out of 614 EPs screened, we encountered a phenotype on their own in 2.93% and an absence of interaction in 96.74% of them. We identified two lines (P{XP}CG8316^d01774 and P{EPgy2}EY06842, referred as d01774 and EY06842 in the text) that specifically enhanced the YARS_E196K-associated rough eye phenotype (Fig. 3I and 3O). Expression of d01774 alone had mild effect on the ommatidial organization (Fig. 3H), which was not further modified upon co-expression of YARS_Wt (Fig. 3J). GMR-Gal4-driven expression of EY06842 did not exhibit phenotype per se, nor when co-expressed with YARS_Wt (Fig. 3N and 3P).

We further tested whether the mutant YARS-specific modifiers alter the GARS_P234KY induced eye phenotype. We co-expressed GARS_P234KY with d01774 or EY06842 and we observed a strong enhancement in both cases (Fig. 3K and 3Q). The co-expression of GARS_Wt with the modifiers did not exhibit any phenotype, confirming the mutant-specificity of the interactions (Fig. 3L and 3R).

We aimed to identify the genes targeted by the d01774 and EY06842 elements. The insertion site of all EP lines used in our screen has previously been molecularly mapped (Bellen et al., 2004; Thibault et al., 2004). EY06842 was annotated as inserted in CG43132 (Drosophila genome release 5.33) and ~1kb upstream from CG15599; the insertion of d01774 was in the 3^rd exon of CG8316 We have remapped the insertion sites of both EP-elements and confirmed their molecular coordinates and orientation (Supplemental Fig. 3A and 3B). EY06842 could either disrupt the CG43132 gene, or promote transcription of the neighboring CG15599 gene. An independent EP-element inserted in CG43132 (P{EPgy2}EY02909) in the 5′-3′ orientation, whose topology was re-confirmed (data not shown), did not interact with YARS_E196K and GARS_P234KY alleles. This excluded the possibility that alteration of CG43132 expression by the EP-element is causing the rough eye enhancement. Furthermore, by co-expressing an RNAi transgene against CG15599 (GMR-Gal4>UAS-YARS_E196K>EY06842>RNAi^CG15599), the rough eye phenotype was fully suppressed (Fig. 3S), confirming the modifier identity. Similarly, downregulation of CG8316 transcript using RNAi reverted the rough eye phenotype of GMR-Gal4>UAS-YARS_E196K>d01774 flies and proved that the gene targeted by d01774 is CG8316 (Fig. 3M). Notably, the YARS protein levels were not changed in the background of d01774 or RNAi^CG8316, indicating that the modification effect was not due to altered YARS_E196K expression (Supplemental Fig. 2C).

Furthermore, we overexpressed d01774 and EY06842 in the retina of two fly neurodegenerative models (Gl-Tau and Gl¹) (Jackson et al., 2002; Fan and Ready,1997). Pathological alterations in Microtubule-associated protein tau are implicated in several brain neurodegenerative disorders, including Alzheimer’s disease and frontotemporal dementia (Spillantini and Goedert, 2013). Mutations in the human ortholog of Gl¹ (Dynactin 1) are associated with motoneuron disorders like distal hereditary motor neuropathy with vocal paresis (Puls et al., 2003) and amyotrophic lateral sclerosis (Munch et al., 2004; Munch et al., 2005). The rough eye phenotypes of Gl-Tau and Gl¹ flies were not enhanced by GMR-Gal4 induced expression of d01774 or EY06842, emphasizing their specific interaction with mutant forms of YARS/GARS, rather than enhancing additive effects unrelated to the aaRS pathomechanism (Supplemental Fig. 3C–3K).

An initial database search revealed that the common YARS and GARS modifiers have low to moderate expression throughout the developmental stages of Drosophila and interestingly, both genes share similar expression dynamics (modENCODE expression cluster 21) (Roy et al., 2010). They encode orphan proteins with unknown molecular function that have orthologues only within the Drosophila genus (12 and 5 orthologues for CG8316 and CG15599, respectively) (Supplemental table 2A). Neither a conventional BLAST, nor remote homology searches could identify other convincing orthologues outside the genus, as the hits had a very low sequence identity and were very diverse in function (Supplemental table 2B). Domain analysis detected hits to part of cyclin dependent kinase 7 (CDK7) and Med15, but these are likely artefactual as they are not conserved even in the orthologues of the modifiers in other Drosophila species (Supplemental figure 4). The crystal structure of both proteins or any close orthologues is unknown. In the N-terminal part of CG8316 a coiled-coil structural motif is predicted, while the rest of the protein is predicted to be largely disordered (Fig. 4A). The CG15599 protein is largely disordered; secondary structures are predicted mainly near the C-terminus (Fig. 4B). Feature-based function prediction indicated that both proteins are likely involved in the regulation of transcription and cytoskeletal protein binding (Supplemental Table 3). Furthermore, both proteins contain predicted nuclear localization sequence (NLS) motifs, CG15599 has a predicted nuclear export sequence (NES) motif (Fig. 4A and 4B), and their nuclear localization is computationally predicted by three different programs (Supplemental Table 4). Thus, extensive data mining could not identify vertebrate orthologues, however it provided potentially useful predictions as to the structure and function of CG8316 and CG15599.

Fig. 4. Schematic representation and disorder propensities of CG8316 (A) and CG15599 (B).

The Y-axis represents the disorder tendency values per amino acid residue as calculated by the IUPred program run on the ELM server. The X-axis depicts the number of amino acid residues. The dashed line at 0.5 of Y-axis indicates the threshold value for disordered/structured residues. Scores above 0.5 correspond to the disordered residues/regions, whereas scores below 0.5 indicate residues/regions predicted to be ordered. Black boxes in (A) and (B) indicate nuclear localization signal motifs (NLS), patterned box - nuclear export signal motif (NES), and CC - coiled-coil domain as predicted by ELM. CG8316 has two predicted overlapping NLS motifs with probability scores p = 0.00072 and p = 0.0013 respectively. CG15599 has NES (p = 7.62e⁻⁰⁴) and NLS (p = 1,276e⁻⁰³) motifs.

Discussion

Glycyl-tRNA synthetase is the most studied of the aaRSs involved in CMT, however molecular and cellular mechanisms linking this and the other aminoacyl-tRNA synthetases to PNS degeneration remain unclear (Motley et al., 2010). Mechanistic insights into CMT2D pathology came from cellular and yeast studies (Antonellis et al., 2006; Nangle et al., 2007), as well as 2 mouse models having motor and sensory deficits (Achilli et al., 2009; Seburn et al., 2006). Two CMT2D mutations (E71G and L129P) were examined in fruit flies, but no phenotypes could be detected even in aged flies (Chihara et al., 2007). Therefore, our study reports the successful generation of a CMT2D Drosophila model, where the effect of two GARS mutations was studied at behavioral, electrophysiological and morphological levels, and upon ubiquitous or spacio-temporally restricted expression. In addition to the patient mutation G240R, we studied the P234KY mutation in the fly for three reasons. First, we opted to compare mutations affecting neighboring residues in the catalytic domain of GARS, with divergent effects on charging function (fully active and completely inactive, respectively). Second, we can benefit from the advantages offered by Drosophila as a model organism to gain mechanistic insights and identify therapeutic targets, which can be further validated in the existing mouse model. Finally, the phenotype of Gars^P234KY mouse is arguably more severe than any of the human disease phenotypes and most likely to have deleterious effects in vivo.

Unlike the YARS-encoded protein which charges tRNA only in the cytoplasm, GARS encodes both the cytoplasmic and mitochondrial isoforms of glycyl-tRNA synthetase. In order to compare the effects induced by both synthetases, we expressed only the cytoplasmic form of GARS in the flies. Based on our experience with the DI-CMTC model (Storkebaum et al., 2009), we tested the Wt and mutant animals in the same experimental paradigms. Similar to YARS_E196K (Storkebaum et al., 2009), ubiquitous expression of GARS_P234KY severely affected the fly fitness with no adults emerging. Since both mutations do not impair the aminoacylation activity of the respective proteins and they are overexpressed on the background of active endogeneous enzymes, these results reconfirm that loss of canonical activity is not a prerequisite for aaRSs toxicity. Ubiquitous expression of the enzymatically inactive GARS_G240R protein had a dosage-dependent effect on the preadult lethality, as has been observed for the YARS mutants with reduced or fully abolished aminoacylation activity (YARS_156VKQV159 and YARS_G41R) (Jordanova et al., 2006; Storkebaum et al., 2009). The neurotoxic effect of GARS_P234KY expression was further demonstrated at the panneuronal level, where again no adult flies emerged. Panneuronal expression of GARS_G240R rendered flies exhibiting progressive motor impairment. These findings further underscore the important role of GARS for normal neuronal function, which is abolished to a different degree by the two mutations.

When targeted to the giant fiber circuit, the GARS_G240R flies only mildly impaired the strength and reliability of the GF-TTMn synapse. Contrastingly, expression of GARS_P234KY during GF development fully disrupted, or severely weakened, the function of the GF to TTMn connection, which was associated with increased RL and a compromised ability to follow high frequency stimulations, similar to DI-CMTC flies (Storkebaum et al., 2009). Consistent with the electrophysiological phenotype, the synaptic terminal of the GF in GARS_P234KY animals was abnormally thin. This suggests that GARS_P234KY interferes with the development of the giant axon terminal. In addition, we observed numerous large vacuoles in the axon and synaptic terminal of GARS_P234KY but not GARS_Wt animals. Although the nature of the vacuoles remains to be determined, they may be indicative of degenerative phenotype. This is supported by the finding that expression of GARS_P234KY after the GF-TTMn connection was established, also disrupted the function of the GF synapse. In addition, expression of GARS_P234KY in the TTMn affected the function of the GF-TTMn connection although much less severe, suggesting that mutant GARS has also a toxic effect in the motoneuron. Therefore, it seems plausible that GARS_P234KY interferes with neuronal differentiation and maintenance of the electrical-chemical GF synapse presynaptically and postsynaptically, but the exact role of GARS in these processes remains to be elucidated. Finally, the anatomical and functional GF phenotypes are similar to those previously described for the expression of DI-CMTC mutations in YARS (Storkebaum et al., 2009), suggesting that mutant YARS and GARS disrupt neuronal development, maintenance and function via similar mechanisms.

The enzymatically active GARS_P234KY protein induced more severe defects in all paradigms. This nicely correlates with what is described in mice. The Gars^P234KY, when compared to Gars^C201R mouse, has more severe neuronal defects, and exhibit decreased life span and body mass (Motley et al., 2010). The deleteriousness of this particular mutation is therefore well conserved between flies and mice, highlighting the complementarities of both models.

Furthermore, we detected similar patterns of toxicity in the CMT2D and DI-CMTC flies, suggesting that the nervous system of Drosophila is susceptible to the effects induced by two CMT-associated aaRSs. Noteworthy, expression of the cytoplasmic isoform of mutant GARS was sufficient to induce phenotypes. This suggests that neurodegeneration occurring in both CMT subtypes can be triggered by pathways outside the mitochondria. One approach to identify those pathways is by identifying the interacting partners of the mutant enzymes. Since our fly models faithfully recapitulate the disease pathophysiology, they could be used for further interaction studies.

The identification of genetic modifiers of CMT-associated genes has not previously been performed in Drosophila. Here, in the context of our DI-CMTC and CMT2D models we initiated an unbiased pilot screen using the rough eye phenotype common to mutant alleles of both aaRSs. Our genetic screen identified CG8316 and CG15599 as shared mutant-specific enhancers. Further studies of the modifiers were compromised by their unknown molecular function and lack of orthologues outside the Drosophila genus. Extensive in silico domain and structural analysis on these orphan proteins could not provide direct insight into their function. Coiled-coil structures, as present in CG8316, are found in proteins with very diverse functions. Unstructured regions, as annotated in CG8316 and CG15599, are also not immediately functionally informative. However, a feature-based machine-learning approach could assign regulation of gene expression and interaction with cytoskeletal proteins as common functions for CG8316 and CG15599. Both modifier proteins were also predicted to be located in the nucleus by several computational methods. Interestingly, many aaRSs (including YARS) are detected in the nucleus of eukaryotic cells (Nathanson and Deutscher, 2000; Popenko et al., 1994; Fu et al., 2012). Their currently known nuclear functions are related to quality control of tRNA nuclear export and transcriptional regulation of protein-coding and ribosomal genes (Grosshans et al., 2000; Ko et al., 2000; Lund and Dahlberg, 1998; Sarkar et al., 1999; Yannay-Cohen et al., 2009). The predicted functions of the modifiers could point to a putative impairment of these biological processes in the aaRSs-related neurodegeneration, however additional studies are necessary to confirm this possibility. Furthermore, identification of more genetic modifiers will enrich the aaRSs interaction network and will provide supporting evidence for the abovementioned or new molecular pathways to be tested experimentally in neuron-specific functional assays.

Our results represent the first genetic evidence that two genes interact in common with CMT-causing alleles of both GARS and YARS. In the context of a large-scale screening strategy in Drosophila, these findings are encouraging as we anticipate being able to pinpoint additional commonalities between the YARS- and GARS-related toxicity. We hypothesize that the targets of these mutant proteins will likely be shared, and analyzing two disease models will help eliminate targets that are not central to their PNS pathogenesis. In the context of CMT disease, putative common pathomechanisms would unravel avenues for targeted drug design and therapy for both CMT forms and possibly other aaRS-related neuropathies.

Highlights.

• First animal model of a GARS mutation causing peripheral neuropathy in humans

• CMT mutations in GARS and YARS induce comparable phenotypes in Drosophila

• The cytoplasmic isoform of GARS is sufficient to trigger neurotoxicity in flies

• CMT mutations in YARS and GARS have common genetic modifiers in Drosophila