Mutations in glycyl-tRNA synthetase impair mitochondrial metabolism in neurons

Abstract The nuclear-encoded glycyl-tRNA synthetase gene (GARS) is essential for protein translation in both cytoplasm and mitochondria. In contrast, different genes encode the mitochondrial and cytosolic forms of most other tRNA synthetases. Dominant GARS mutations were described in inherited neuropathies, while recessive mutations cause severe childhood-onset disorders affecting skeletal muscle and heart. The downstream events explaining tissue-specific phenotype–genotype relations remained unclear. We investigated the mitochondrial function of GARS in human cell lines and in the GarsC210R mouse model. Human-induced neuronal progenitor cells (iNPCs) carrying dominant and recessive GARS mutations showed alterations of mitochondrial proteins, which were more prominent in iNPCs with dominant, neuropathy-causing mutations. Although comparative proteomic analysis of iNPCs showed significant changes in mitochondrial respiratory chain complex subunits, assembly genes, Krebs cycle enzymes and transport proteins in both recessive and dominant mutations, proteins involved in fatty acid oxidation were only altered by recessive mutations causing mitochondrial cardiomyopathy. In contrast, significant alterations of the vesicle-associated membrane protein-associated protein B (VAPB) and its downstream pathways such as mitochondrial calcium uptake and autophagy were detected in dominant GARS mutations. The role of VAPB has been supported by similar results in the GarsC210R mice. Our data suggest that altered mitochondria-associated endoplasmic reticulum (ER) membranes (MAM) may be important disease mechanisms leading to neuropathy in this condition.


Introduction
All genes are copied into short-lived RNA molecules, which are then translated to proteins, forming the building box of the cells in the body. Although the majority of protein synthesis happens in the cytosol, an additional translation apparatus is required to translate 13 proteins important in mitochondrial energy production, which are encoded by the mitochondrial genome (1). The majority of genes which regulate protein translation in these cellular compartments are distinct, but two genes encoding aminoacyl-tRNA synthetases of glycine (GARS) and lysine (KARS) are common, in both mitochondria and the cytosol (2). Investigating the function of these bi-functional enzymes provides a unique opportunity to explore interactions between cytosolic and mitochondrial translation and facilitates the dissection of tissue-specific manifestations of mitochondrial or cytosolic defects of translation.
GARS encodes the non-redundant homodimeric glycyl-tRNA synthetase, covalently attaching glycine to its cognate tRNA, which is essential for the fidelity of protein translation (2). There are two translation initiation sites resulting in the production of mitochondrial and cytoplasmic isoforms of GARS. Three functional domains are common in both isoforms: a highly conserved N-terminal WHEP-TRS domain, a catalytic core and a C-terminal anticodon-binding domain, however a mitochondrial targeting signal (MTS) is only present in mitochondrial GARS (Fig. 1A). Autosomal-dominant GARS mutations cause axonal CMT (CMT2D) or distal hereditary motor neuropathy with upper limb predominance (dHMN-V) (2). While the majority of CMT2D mutations were localized to the catalytic domain, autosomal recessive mutations were reported in three independent patients with a mitochondrial phenotype (Fig. 1A). One baby boy, homozygous for c.2065C>T, p.(Arg689Trp), had severe neonatal cardiomyopathy and cytochrome c oxidase deficiency and died at 10 days of age (3). Another child with compound heterozygous c.1904C>T, p.(Ser635Leu) and c.1787G>A, p.(Arg596Gln) presented with exercise-induced myalgia, non-compaction cardiomyopathy, periventricular lesions and increased lactate (4). Recently, recessive mutations within the catalytic domain were also reported causing multisystem disease with growth retardation, delayed motor milestones, dysmorphic signs and complex neurological presentation of microcephaly, thinning of the corpus callosum, white matter lesions, cerebellar vermis and brainstem atrophy, but without peripheral neuropathy (5). To date no neuropathy was observed in children with recessive mutations however it may develop later in life. A mild neuropathy was observed on electrophysiological testing in the heterozygous father of the second child (4).
Reduced aminoacylation activity, altered axonal localisation (6,7), impaired catalytic function (8) and more recently the altered neuropilin 1 pathway (9) were described to contribute to the disease in GARS mutations. However, to date, little is known about the mitochondrial role of GARS and how it affects the disease phenotype. Thus, in this study, we explored the mitochondrial function of the bi-functional enzyme and show that mutations in GARS lead to tissue-specific mitochondrial defect in neurons with different pathomechanism in autosomal dominant and recessive mutations.

Patients
We studied cell lines of patients carrying pathogenic GARS mutations ( Fig. 1B and C). Patient 1 and his mother (II/1) manifested with typical clinical and electrophysiological signs of dHMN-V, both developed upper limb predominant distal neuropathy since early twenties (Fig. 1B). They both carry the c.647A>G, p.(His216Arg) mutation in GARS, detected on a diagnostic multigene next generation panel. This variant is predicted to be pathogenic based on Association for Clinical Genetic Science (ACGS) Practice Guidelines (2013) and segregated with the phenotype in the family The clinical presentation and family history of patient 2 have been reported previously (4). The 12-year-old girl developed cardiomyopathy and exercise intolerance at age 6 and carries the compound heterozygous c.1904C>T, p.(Ser635Leu) and c.178G>A, p.(Arg596Gln) GARS mutations. The mother is heterozygous for the c.1787G>A, p.(Arg596Gln) mutation and shows no signs of a neuropathy, the father has a mild neuropathy and carries c.1904C>T, p.(Ser635Leu).

GARS is localized to mitochondrial RNA granules
To prove that GARS is present in the mitochondria and is involved in mitochondrial translation, we performed double immunofluorescence staining (MitoTracker Red and GARS) in human fibroblasts ( Fig. 2A-C) and oligodendroglia cells ( Fig. 2E-G). GARS was detectable both in the cytoplasm and mitochondria ( Fig. 2A-L) and was present in mitochondrial RNA granules ( Fig. 2M and N), which was confirmed using multiple markers including Mitotracker, MTCO1 and GRSF1, respectively. GARS-associated aggregates in fibroblasts were previously reported (10). These cytoplasmic granules labelled with eIF4E representing processing bodies (PB) or stress granules (SG) did not confirm co-localization ( Fig. 2O and P). These results show that GARS is present in RNA granules in mitochondria, thus may affect mitochondrial translation.
siRNA down-regulation of GARS results in abnormal mitochondrial translation in myoblasts and oligodendroglia We first tested in vitro whether mitochondrial protein synthesis in fibroblasts is altered by any of these variants. None of the fibroblasts from these patients showed a mitochondrial translation defect, which is not uncommon in mitochondrial diseases with tissue-specific clinical presentations. Next we performed siRNA-mediated downregulation of GARS with two different siRNAs in normal human fibroblasts, myoblasts and also in SHSY-5Y cells, providing the most relevant cellular model of GARS-related disease. We detected significantly decreased protein levels of respiratory chain subunits in myoblasts ( Fig. 2Q and S) and SHSY-5Y cells ( Fig. 2Q and T), but not in fibroblasts ( Fig. 2Q and R), which is illustrating that loss of function of GARS result in a tissue-specific mitochondrial translation defect. Both siRNAs successfully down-regulated GARS in all cell types both in the cytosol and mitochondria (data not shown). heterozygous parents of patient 2 (carrier 1 and 2) whose fibroblasts were used in this study.

Reduced mitochondrial respiratory chain enzymes in induced neuronal progenitor cells (iNPCs) of patients carrying dominant and recessive GARS mutations
Based on the neuron-specific manifestations in patients, we studied neuronal cell lines. We took advantage of a recently developed method converting human fibroblasts directly into iNPCs to model neurodegenerative disorders (11). Within one week of conversion, the expected change in morphology was observed (Fig. 3A) and efficient viral transduction was confirmed by quantitative RT-PCR (Supplementary Material, Fig. S1A-D). Primary fibroblasts from two patients, two GARS mutation carriers and two healthy controls were converted (Fig. 3B) and RT-PCR ( Fig. 3B-D), immunofluorescence staining (Fig. 3E) and western blot (Fig. 3F) for fibroblast and iNPC markers revealed successful down-regulation of fibroblast markers and strong upregulation of iNPC markers in all cell lines (Supplementary Material, Fig. S1E).

RNA sequencing (RNAseq) in induced neuronal progenitor cells (iNPCs) of patients 1 and 2
To further address the conversion toward neuronal cell linage, RNA isolated from control fibroblasts and control and patient iNPC cells, was subjected to RNAseq analysis. The iNPC cell lines of patient 1, patient 2 as well as a control iNPC line were clearly grouped together in a hierarchical clustering of the RNA read count sample distance matrix (Fig. 4A). This indicates that the iNPCs are not related to fibroblasts and they are likely to exhibit a functionally distinct and unique transcriptome. Examination of the log 2 fold change in control iNPC versus control fibroblast RNAseq read counts revealed that thousands of genes exhibited a significant difference in expression between the fibroblast and iNPC lines (Fig. 4B). This included differential expression of genes involved in neuronal function. GO terms that were significantly enriched for genes showing significant up-/down-regulation of RNA expression in both patient iNPCs lines versus control iNPCs are represented in Figure 4C, compared with the distinct lists of the most significantly enriched GO terms, specific to each patient shown in Supplementary Material, Figure S2A and B. In further support of the neuronal fate in our iNPCs neuronal markers NES, CD24, ROR2, FZD3, USP44 are significantly overexpressed in iNPCs compared with fibroblasts and we detected SOX2 expression, which is not present at all in fibroblasts. These neuronal markers have been previously reported as specific for neuronal stem cells and provide evidence of a neuronal fate in our iNPCs (12). We could not detect significant changes in mRNA expression of genes of mitochondrial proteins which showed significant changes in the proteomic data, therefore we suggest that the changes occur on the protein level, which is in line with the defect in protein translation and altering protein-protein interactions.
After confirming the successful conversion of the cell lines, we measured mitochondrial function in iNPCs. The number of mitochondria and the copy number of mtDNA gradually increased in iNPCs and the defect became more pronounced after eight passages, confirming the higher energy demand in neuronal cells (Supplementary Material, Fig. S3A). While no defect was observed in fibroblasts ( Fig. 5A and B, E and F), we detected significantly decreased level of the complex I subunit NDUFB8 in patient 2 and the mildly affected father (carrier 2), while a mitochondrial encoded complex IV subunit (MTCO1) was significantly reduced in iNPCs of patient 1 ( Fig. 5C and D). Mitochondrial proteins remained normal in control iNPCs and in the healthy mother (carrier 1) carrying the heterozygous p. (Arg596Gln).
BN-PAGE showed significantly lower levels of all five mitochondrial complexes in iNPCs of patient 1, but no defect was observed in any of the other iNPC lines ( Fig. 5G and H) or in fibroblasts ( Fig. 5E and F). Interestingly, mitochondrial GARS was increased in symptomatic patients' fibroblasts, suggesting compensation, possible through redistribution (as seen in mice), while in iNPCs from all patients and carriers, the amount of GARS was consistently lower in both mitochondria and cytosol (Supplementary Material, Fig. S3B-E). Immunofluorescent staining confirmed lower expression of respiratory chain subunits of complexes I and IV (NDUFS2, MTCO2) in iNPCs of all symptomatic individuals (Fig. 6) raising the possibility that the mitochondrial translation defect may be caused by the loss of function of GARS in neurons.
MitoTracker Red CMXRos accumulates only in actively respiring mitochondria that have an intact membrane potential (DWm) (13). A deficient MitoTracker Red staining and the less elongated appearance of mitochondrial networks in iNPCs derived from patient 1 suggest that DWm is altered in these cell lines (Fig. 6).
Mitochondrial respiration measured on Seahorse XF96 showed that all fibroblasts from affected individuals with different GARS mutations had slightly higher maximal respiration than controls, suggesting compensation, while there was no difference in basal respiration and proton leak ( Fig. 7A-D). However, we detected significantly reduced respiration in iNPCs of patient 1 ( Fig. 7A-D). Additional analysis of iNPCs of patient 1 revealed increased oxidative stress (CellROX V R Green staining) ( Fig. 6E and F).
In summary, we detected a neuron-specific defect of mitochondrial proteins in iNPCs of patient 1, which correlated with the neuropathy.

Comparative mitochondrial proteomics identified different alterations in iNPCs of patients with dominant and recessive GARS mutations
Comparative proteome profiling in iNPCs of patient 1 (dominant) and patient 2 (recessive) identified significant changes in several mitochondrial proteins when compared with control iNPCs. A detailed analysis of cellular processes influenced by these proteins was performed using STRING (Search Tool for RNA granule marker GRFS1 (red) and GARS (green) in fibroblasts (M). Boxed area is shown as zoomed image (N). Staining for cytosolic RNA granule marker eIF4E (red) and GARS (green) in fibroblasts (O). Boxed area is shown as zoomed image (Q). Human fibroblasts, myoblasts and SHSY-5Y cells were subjected to siRNA mediated GARS downregulation. Cells were transfected with non-targeting (NT) siRNA, and two different GARS siRNAs (GARS siRNA 1 and 2) for 9 days. After 9 days of treatment total cell lysates were collected from the cells. The silencing of GARS protein and the steady-state level of mitochondrial proteins were determined by western blots.
GAPDH was used as loading control (R, S, T). Bar graph indicates the level of mitochondrial proteins in un-transfected (control), NT siRNA transfected and GARS siRNA treated fibroblasts (R), myoblast (S) and SHSY-5Y cells (T). Error bars show standard deviation. Three independent experiments were performed. One-way ANOVA followed by Bonferroni post-hoc test (with 97% confidence intervals) were used in comparing the level of protein expression (*P < 0.5 versus cont. and non-targeted samples). the Retrieval of Interacting Proteins) software (version 10.0). The protein-protein interactions between the altered proteins analysed by STRING is shown in Figure 8A and B. We observed >2 times different abundance in both patients' iNPCs of subunits and assembly factors of respiratory chain enzyme complexes, Krebs cycle enzymes and proteins involved in mitochondrial transport. Significant alterations of fatty acid oxidation proteins were only observed in the mitochondrial patient's iNPCs (Fig. 8A). VAPB and proteins acting downstream from VAPB were significantly down-regulated only in iNPCs of patient 1 (Fig. 8B). VAPB binds to protein tyrosine phosphatase interacting protein 51 (PTPIP51), and is known to be involved in connecting ER with mitochondria as part of the mitochondria-associated ER membranes (MAM) complex (14). Based on the differentially expressed proteins we identified several biological processes different in the mitochondrial or neuropathy cell lines (Supplementary Material, Table S1).

Mitochondrial calcium uptake was significantly altered in fibroblasts and iNPCs of patients carrying pathogenic GARS mutations
To confirm the altered expression of VAPB in patient cells we performed immunoblotting for VAPB and its interacting partner PTPIP51 in both fibroblasts and iNPC cells ( Fig. 9A and B). Our data indicated significantly reduced levels of VAPB and PTPIP51 in the cells of patient 1, and more importantly the lack of VAPB protein was more evident in the iNPCs compared to fibroblasts (Fig. 9B).
The primary role of MAM is to facilitate delivery of Ca 2þ to the mitochondrial matrix from ER stores. Thus, reduced levels of VAPB loosen the ER-mitochondrial associations and perturb this delivery. To address the possible role of GARS in mitochondrial Ca 2þ homeostasis we monitored the mitochondrial Ca 2þ uptake following its release, induced by histamine, from ER stores. For these experiments we used both fibroblasts and iNPCs cells. After recording the baseline where indicated ( Fig. 9E and G), cells were exposed to 100 lM histamine (His), causing the generation of inositol 1, 4, 5 trisphosphate (InsP 3 ) and the consequent opening of the InsP 3 channels of the intracellular stores. Representative traces of typical experiments for fibroblasts and iNPCs are shown in Figure 9E and G. Mitochondrial Ca 2þ influx in cells of patient 1 were significantly lower in both cell types than those observed in control or cells of patient 2. Peak values are shown in Figure 9F and H.
To complement these studies, we also analysed the autophagy in the iNPC cells. Recently published work showed that the VAPB-PTPIP51 tethers regulate autophagy by mediating the delivery of Ca 2þ to the mitochondria from the ER stores (15). Autophagy detection kit was used specifically label autophagosomes (Fig. 9I). Fully in line with our findings we identified significantly enlarged and increased numbers of autophagic vacuoles in the neuropathy cell line compared with control cells (Fig. 9J and K).

Investigations in the Gars C210R mice detected significant changes in MAM proteins and mitochondrial dysfunction in sciatic nerve
To investigate the effect of GARS mutations on mitochondrial translation and on MAM proteins we studied different tissues in Gars C201R mice. Homozygous Gars C201R/C201R mice have severe motor deficits and die between 14 and 17 days. Heterozygous Gars C201R/WT mice present with a reduction in axon diameters and muscle weakness at 3 months of age (16). First we studied mitochondrial protein synthesis by 35 S methionine pulse-labelling in primary fibroblasts isolated from wild-type, heterozygous and homozygous mice. Our data, using Coomassie blue stained gel as loading control, showed normal result (Supplementary Material, Fig. S6). We could not detect a mitochondrial translation defect in mouse fibroblasts, similar to our data in patient fibroblasts.
Next, we investigated the expression level of VAPB and PTPIP51 in isolated sciatic nerve at P7. We detected significantly reduced levels of VAPB and its interacting partner PTPIP51 in homozygous Gars C201R/C201R mice while no changes were found in heterozygous Gars C201R/WT mice (Fig. 10A, D and E). Furthermore, detection of GARS protein revealed a complete lack of gars dimers in Gars C201R/C201R animals which are the active form of the enzyme [ Fig. 10A (upper band) and B]. On the other hand the level of GARS monomer was unaltered compared to wild-type littermates [ Fig. 10A (lower band) and C]. Based on our human iNPCs data indicating mitochondrial defect in patients, we also studied the level of mitochondrial proteins in different mouse tissues. The amount of mitochondria detected by porin expression suggested significant decrease ( Fig. 10A and F) in the sciatic nerve of the Gars C201R/C201R animals and we detected significantly lower level of complex I, IV and V subunits ( Fig. 10G and H). Other highly metabolic tissues were also analysed (skeletal muscle, brain, heart, liver and kidney) at In summary, extensive investigations in the Gars C210R mice detected tissue-specific mitochondrial dysfunction and altered level of MAM proteins in sciatic nerve only, while all other tissues showed normal results.

Discussion
It has been shown in mice that mutant GARS acquires a neomorphic binding activity to neuropilin 1 (Nrp1) receptors that directly antagonizes an essential signalling pathway for motor neuron survival (9). This aberrant interaction interferes with the binding of the vascular endothelial growth factor (VEGF) to Nrp1. Genetic reduction of Nrp1 in mice worsens neuropathy, whereas enhanced expression of VEGF improves motor function. It seems to be an important pathomechanism of the secreted GARS outside the neurons however it does not explain all abnormalities observed in GARS mutations.
In this study we identified that reduction or loss of GARS (by siRNA down-regulation or some pathogenic mutations) result in decreased translation of mtDNA and nuclear encoded subunits confirming that GARS is important for obtaining mitochondrial   function, and that loss of function of GARS result is reduced mitochondrial translation. We showed that siRNA downregulation of GARS resulted in reduced mitochondrial translation in neuronal cells and myoblasts, but not in fibroblasts, suggesting that GARS mutations may lead to a tissue-specific mitochondrial translation defect due to loss of function. We could not demonstrate a defect of mitochondrial protein synthesis in a mouse model of GARS (Gars C201R ) in five different tissues. The steady-state level of Gars was not altered by the mutation, and we detected redistribution of Gars into mitochondria, which may have prevented respiratory chain dysfunction, as higher protein levels of Gars were detected in mitochondria in the studied tissues.
Furthermore, we detected complex alterations of mitochondrial metabolism, which cannot be fully explained by a defect of cytosolic and mitochondrial translation suggesting that GARS has other non-canonical functions within neurons (17,18). The canonical role of GARS would be the addition of glycine residues to the cognate tRNA during mitochondrial and cytosolic protein translation and we focussed on studying mitochondrial translation. We detected partial co-localization of GARS with mitochondrial RNA granules. Mitochondrial RNA granules are the centres of posttranscriptional RNA processing and the biogenesis of mitochondrial ribosomes (18) and provide a platform for the spatiotemporal regulation of numerous processes including RNA maturation, ribosome assembly and translation initiation (19), supporting the role of GARS in mitochondrial translation and raising the possibility that mutant GARS may cause alteration of these functions. However mitochondrial RNA granules are not specific for neurons and do not explain the tissue specificity in the pathomechanism of GARS mutations.
We generated iNPCs by direct conversion from human fibroblasts and show that this is an excellent cellular model to study tissue-specific mechanisms in neurons. We detected an abnormal translation of mtDNA and nuclear-encoded mitochondrial proteins and abnormal mitochondrial respiration in neuronal cells, but not in fibroblasts of patients compared with controls. Interestingly, these changes were more prominent in iNPCs of the dHMN-V patient carrying a dominant, probably gain-offunction GARS mutation (patient 1), suggesting that some neuropathy-associated mutations result in more severe or complex alteration of mitochondrial function in neurons.
We further investigated the mitochondrial defect by RNAseq and by comparative proteomics in iNPCs carrying autosomal dominant and recessive GARS mutations which showed several significant changes of the mitochondrial transcriptome and proteome. Altered levels of mitochondrial respiratory chain subunits, Krebs cycle enzymes and proteins involved in mitochondrial transport and respiratory chain assembly were detected in both mutant iNPCs. In addition, significant changes were identified in the fatty acid metabolism only in iNPCs with recessive GARS mutations (Fig. 8A). In support of the relevance of these findings, exercise intolerance, myopathy and cardiomyopathy, typically present in defects of fatty acid oxidation, were observed in the patients reported to date with recessive GARS mutations (20).
iNPCs carrying a neuropathy-related dominant GARS mutation showed reduced VAPB and its downstream pathways, including altered mitochondrial and cellular calcium homeostasis, providing a potential explanation for neuron-specific clinical manifestations (Fig. 8B). Autosomal dominant mutations in VAPB were reported in several families world-wide with clinical presentations of ALS (ALS8) or distal spinal muscular atrophy (14,21). VAPB is part of the mitochondria-associated ER membrane (MAM) complex and was shown to bind to the outer mitochondrial membrane protein PTPIP51 to tether ER to mitochondria and plays an important role in the quality control and provides an additional level of flexibility of neurons to cope with misfolded protein stress in the ER (22). Modulating VAPB induces changes in ER-mitochondria contacts and Ca 2þ exchange between the two organelles, and predicted to lower mitochondrial ATP production (22). It has been very recently shown that the VAPB-PTPIP51 tethers regulate autophagy through mediating delivery of Ca2þ to mitochondria from ER stores, providing a new molecular mechanism for regulating autophagy (15). In support of the relevance of reduced VAPB, we also detected low protein levels of its direct interacting partner PTPIP51 and several proteins involved in pathways downstream of VAPB were altered including calcium metabolism, ubiquitination, ephrin signalling, ER trafficking and vesicle fusion (SNARE) (Fig. 8B).
In support of the functional effect of reduced VAPB, we detected a significant defect of mitochondrial calcium metabolism in fibroblasts and neuronal cells in the dHMN-V patient (patient 1). Furthermore, significant increase of autophagy was observed in the patient cells ( Fig. 9I and J) and the average size of autophagic structures was significantly increased specifically in patient 1 (Fig. 9K). Furthermore, extensive investigations in the Gars C210R mice detected tissue-specific mitochondrial dysfunction and altered level of MAM proteins in sciatic nerve only, while all other tissues showed normal results. Based on the reduced level of MAM proteins and respiratory chain subunits we think that the homozygous p.Cys201Arg mutation has a loss-offunction effect in Gars C201R mice, while we cannot provide experimental evidence of the loss of function in the heterozygous animals.
Although further studies are needed to explore the exact mechanism of the GARS/VAPB connection we suggest that mutant GARS has an impact on mitochondrial calcium metabolism and ER-mitochondria interactions which contribute to the neuron-specific clinical presentations. Furthermore, the changes detected in our study may provide a functional link between the previously identified presynaptic defects of neuromuscular transmission and GARS mutations, as calcium signalling is important for vesicle refinement at the neuromuscular junction (23).
The importance of MAM in motor neurons has been highlighted by the fact, that mitofusin 2 (MFN2), a common cause of axonal CMT (CMT2), has been suggested to act as an ER-mitochondria tether, and its ablation decreases interorganellar communication (24). Many functions regulated by ER-mitochondria associations have been shown to contribute to neurodegenerative diseases such as amyotrophic lateral sclerosis/frontotemporal dementia (ALS/FTD), Alzheimer's disease (AD) and Parkinson's disease (PD), however the exact mechanism involved in damage of different neuronal cell types in these diseases needs further investigations. Here we show that the ER-mitochondria axis is altered in GARS-related neuropathy, highlighting that MAM is particularly important in motor neurons.
In summary, we show that a different type of mitochondrial dysfunction underlies autosomal recessive and dominant GARS mutations. While a defect of mitochondrial protein synthesis in recessive GARS-related mitochondrial disease is most likely a result of loss of function, in dominant neuropathy-causing mutations we identified a defect of the MAM complex, suggesting that mitochondria ER interactions and calcium metabolism, possible due to non-cognate function of GARS may potentially explain tissue-specific clinical manifestations in human motor neurons.

Cell culture
Cultured myoblasts were grown in skeletal muscle cell growth medium and supplement mix (PromoCell) supplemented with 10% (v/v) foetal bovine serum (FBS, Sigma Aldrich) and 4-mM Lglutamine (Invitrogen) and cultured as recommended by the supplier. Fibroblasts were grown in high glucose Dulbecco's modified Eagle's medium (DMEM, Sigma, Poole, UK) supplemented with 10% FBS. Myoblasts cell line was immortalized by transduction with a retroviral vector expressing the catalytic component of human telomerase (25). Human oligodendrocyte cells (MO3.13) and a human bone marrow neuroblastoma cell line (SHSY-5Y) were grown in DMEM: F12 supplemented with 10% FBS. The human oligodendroglia cell line was a kind gift of Dr. Josephine Nalbantoglu from the McGill University, where it was originally characterized.

Conversion of skin fibroblasts to iNPCs
The method used was described previously in Meyer et al. (11).

siRNA-mediated down-regulation
Silencer GARS siRNA (number siRNA, Ambion-Life Technologies) was transiently transfected in control myoblasts, fibroblasts and in SHSY-5Y human neuroblastoma cell line at a final concentration of 20 nM using Lipofectamine RNAiMAX (Invitrogen), according to the manufacturer's specifications. Transfections were repeated on day 3 and day 6, cells were harvested on day 9. A non-targeting SilencerSelect Negative Control (#1) was used as a control.

BN PAGE, in-gel activity and immunoblotting
About 3% and 15% gradient gels and a 4% stacking gel were prepared for BN PAGE according to the protocol of Calvaruso et al. (26). A Gilson MiniPuls 3 gradient gel mixer (Gilson, USA) was used at a speed of 5.38 ml/min. SBG buffer [750 mM aminocaproic acid, 5% Coomassie Brilliant Blue G250 (Biorad, UK)) was added to the samples and 2 lg of each sample used. Following electrophoresis the gel was transferred to a PVDF membrane, de-stained, blocked and incubated with primary antibody to the respiratory chain complexes (complex I subunit NDUFA9-2 lg/ml, complex II 70 kDa subunit-0.2 lg/ml, complex III core 1 subunit-1 lg/ml, complex IV subunit 4-1 lg/ml, complex V ATP synthase-0.25 lg/ml, all Abcam, UK), a secondary rabbit antimouse antibody (Dako, UK) (0.5 ll per ml) and then developed with ThermoScientific Pierce ECL2 Western Blotting kit or BioRad Clarity Western ECL Solution. The signal was detected with the UVP BioSpectrum 500 Imaging System and the intensity quantified using Image J software. 'In-gel' assays were carried out as described (27).

Pulse-labelling of mitochondrial translation products
In vivo 35 S-metabolic labelling studies were performed as described previously with the following modifications. Cells, cultured to 60-70% confluency in T25 mm flasks, washed with phosphate-buffered saline (PBS; Sigma) and washed by incubating twice for 10 min at 37 C/5% CO 2 in methionine/cysteine-free DMEM (Sigma, Poole, UK), with the media replaced between each incubation. Cells were then incubated for 15 min at 37 C/5% CO 2 in methionine/cysteine-free DMEM supplemented with 5% (v/v) dialyzed FBS, 0.1 mg/ml emetine dihydrochloride (Sigma). Following addition of 200 mCi/ml 35 S-methionine/cysteine ( 35 S EasyTag EXPRESS; Perkin Elmer), cells were incubated for 15 min at 37 C/5% CO 2 , then washed twice with ice-cold DMEM supplemented with 7.5 mg/ml methionine. Cell pellets were prepared after washing once with ice-cold PBS. Radiolabelled proteins were then analysed using SDS-PAGE as described previously.

Mitochondrial extraction from mouse tissues
Unless specified all chemicals were purchased from Sigma (Sigma Aldrich, UK). Liver, brain kidney heart and muscle tissue was Gene expression studies cDNA was generated by reverse transcription of 500 ng of total RNA using the Superscript VILO cDNA synthesis kit (Invitrogen, 11754-050) according to manufacturers' instructions. qPCR (Applied Biosystems 7900HT) was performed in triplicate on cDNA using SYBR Green PCR Master Mix (Invitrogen, 4309155). Samples were normalized using the average of two reference genes, GAPDH and actin. Primers are shown in Supplementary Material, Table S2.

RNAseq analysis
Total RNA was isolated from control fibroblast cells and from patient and controls iNPC cells using the mirVana TM miRNA Isolation Kit (Ambion) and DNAse treated with the DNA-free TM DNA Removal Kit (Ambion). RNAseq libraries were prepared using Illumina (Illumina, Inc. California, USA) TruSeq Stranded Total RNA with Ribo-Zero Human kit and were sequenced on an Illumina HiSeq 2500 platform using paired-end protocol. The quality of sequencing reads was checked with FastQC (http:// www.bioinformatics.babraham.ac.uk/projects/fastqc/; date last accessed February 20, 2018). Reads were aligned using the STAR (v2.5.2b) aligner and the 2-pass protocol that is outlined in the GATK documentation for calling variants from RNAseq (https:// software.broadinstitute.org/gatk/; date last accessed February 20, 2018). Number of reads mapped to genes were counted using HTSeq-count (30). Differentially expressed genes were identified with Bionconductor package DESeq2 (31), comparing control fibroblast versus iNPC cells, as well as patient iNPC versus control iNPC cells. Genes differentially expressed with a false discovery rate (FDR) of 0.1 and a log 2 fold change ! 1 were considered as differentially expressed genes. Gene set enrichment analysis for gene ontology terms was performed using the CPDB web tool (http://cpdb.molgen.mpg.de/; date last accessed February 20, 2018).

Proteomic analysis: sample preparation
Samples (two biological replicates per condition) were solubilized in 1% SDS (w/v), 50 mM Tris-Cl pH 7.8, 150 mM NaCl, supplemented with protease inhibitor Complete Mini and phosphatase inhibitor cocktail PhosSTOP (Roche, Switzerland). The Bichinonic acid assay (BCA) (Thermo Scientific, USA) was used to determine protein concentrations. Cysteines were reduced with 10 mM dithiothreitol for 30 min at 56 C and alkylated with 30 mM iodoacetamide for 20 min at room temperature, in the dark. Per sample, a volume corresponding to 10 mg of protein was 10-fold diluted with ice-cold ethanol, followed by 60 min of incubation at À40 C. The sample was centrifuged for 30 min at 4 C and 18 000g and the supernatant was discarded. The protein pellet was resolubilized in 5 mL of 4 M guanidinium chloride and diluted 20-fold with 50 mM ammonium bicarbonate (ABC). Trypsin (Promega, USA) was added in a ratio 1: 20 (trypsin: protein, w/w) and samples were digested at 37 C for 14 h. Digestion was stopped by adding trifluoroacetic acid to a final concentration of 1%. Digestions were quality controlled by separation of 1 mg on a monolithic column according to Burkhart et al. (32).

Nano-LC-MS/MS analysis
Per sample 0.5 mg were analyzed using an Ultimate 3000 nano RSLC (Thermo Scientific) coupled to an Orbitrap Fusion Lumos (Thermo Scientific, Germany). Peptides were pre-concentrated on a 75 lm Â 2 cm C18 Acclaim Pepmap trapping column (Thermo Scientific, Germering, Germany) in 0.1% TFA for 10 min using a flow rate of 20 ml/min, followed by separation on a 75 lm Â 50 cm Acclaim Pepmap C18 main column using a binary gradient (A: 0.1% formic acid, B: 84% acetonitrile, 0.1% formic acid) ranging from 3% to 30% B in 85 min at a flow rate of 250 nl/min. Samples were analysed in data-dependant acquisition mode (DDA). MS survey scans were acquired in the Orbitrap from 300 to 1500 m/z at a resolution of 120 000 using the polysiloxane ion at 371.1012 m/z as lock mass (33). For MS/MS precursors were selected using the top speed option (3 s) and a dynamic exclusion of 15 s. Peptides were selected with an isolation window of 1.2 m/z and fragmented using higher-energy collisional dissociation (HCD) with a normalized collision energy of 30%, and fragment ion spectra were acquired in the ion trap in rapid mode. Automatic gain control (AGC) settings were 2 Â 10 5 (MS) and 2 Â 10 3 (MS/MS) ions, maximum injection times were set to 50 ms (MS) and 300 ms (MS/MS), respectively (34,35).

Mitochondrial calcium metabolism studies
Fibroblast and iNPC cell lines were seeded (100 000 cells/dish) on glass bottom dishes and allowed to adhere overnight. Rhod-2am (ThermoFisher) was used to evaluate mitochondrial Ca2þ, respectively. Rhod-2 (8 lM) was loaded at room temperature for 30 min. Cells were washed twice with PBS and incubated in tyrodes solution [137 mM NaCl, 2.7 mM KCl, 1 mM MgCl 2 , 0.2 mM Na 2 HPO 4 , 12 mm NaHCO 3 , 2 mM CaCl 2 , 10 mM Glucose, pH 7 (Sigma-Aldrich) for 15 min]. Dynamic measurements of mitochondrial Ca2þ were performed at room temperature with a Nikon A1R inverted confocal microscope equipped with a Â60, 1.40 NA oil objective in resonant scanning mode at 3.7 fps for 5 min. After 1 min (providing basal fluorescence) histamine (100 lM; Sigma-Aldrich) was added to evoke IP3R-dependent ER Ca2þ release and the Ca2þ transient was monitored. Image analysis was performed in Fiji, the background was subtracted and Ca2þ levels were calculated as (F -F0)/F0 where F indicates fluorescence and F0 basal fluorescence.

Autophagy detection
Control and patient fibroblasts and iNPCs were cultured on coverslips and subjected to autophagy activity assays as described in the manufacturer instructions (Autophagy Detection Kit ab139484, Abcam, Cambridge, MA). For positive control cells were subjected to Rapamycin for 14-16 h.

Proteomic data analysis
Data analysis for label-free quantification was performed with the Progenesis QI from Nonlinear Dynamics (Newcastle upon Tyne, UK). MS raw files were automatically aligned, followed by peak picking. Only peptides with charge states þ2, þ3 and þ4 were considered for peptide statistics, analysis of variance (ANOVA) and principal component analysis (PCA). Only MS/MS spectra with ranks 1-10 were exported as peak lists with a limited fragment ion count 100. Using searchGUI (33) spectra were searched against a target/decoy version of the human Uniprot database (downloaded July 2015, containing 20 207 target sequences) using X! Tandem Vengeance (2015.12.15.2). Additionally, spectra were searched using Mascot 2.5 (Matrix Science). Search parameters were: trypsin as enzyme with a maximum of two missed cleavages, carbamidomethylation of Cys set as fixed modification, oxidation of Met set as variable modification, MS and MS/MS tolerances of 10 ppm and 0.5 Da, respectively. PeptideShaker 1.10.3 was used to combine all search result files and filter data at a FDR of 1% on the protein and peptide level (34). Obtained identifications were exported using the advanced PeptideShaker features that allow direct reimport of the quality-controlled data into Progenesis.
Only proteins quantified with at least one unique peptide were exported from Progenesis. Then, for each protein, the average of the normalized abundances (obtained from Progenesis) from the biological replicate analyses was calculated to determine the ratios between the samples and control. Only proteins which were (i) commonly quantified in all the replicates with (ii) at least two unique peptides, (iii) an ANOVA P-value of <0.05 (Progenesis) and (iv) a 2-fold regulation between samples and controls were considered as regulated.
A set of 85 regulated proteins were used to generate a high confidence STRING network (http://string-db.org/) and to analyse the enrichment of Gene Ontology terms.

Supplementary Material
Supplementary Material is available at HMG online.