Peripheral neuropathy via mutant tRNA synthetases: Inhibition of protein translation provides a possible explanation

Recent evidence indicates that inhibition of protein translation may be a common pathogenic mechanism for peripheral neuropathy associated with mutant tRNA synthetases (aaRSs). aaRSs are enzymes that ligate amino acids to their cognate tRNA, thus catalyzing the first step of translation. Dominant mutations in five distinct aaRSs cause Charcot‐Marie‐Tooth (CMT) peripheral neuropathy, characterized by length‐dependent degeneration of peripheral motor and sensory axons. Surprisingly, loss of aminoacylation activity is not required for mutant aaRSs to cause CMT. Rather, at least for some mutations, a toxic‐gain‐of‐function mechanism underlies CMT‐aaRS. Interestingly, several mutations in two distinct aaRSs were recently shown to inhibit global protein translation in Drosophila models of CMT‐aaRS, by a mechanism independent of aminoacylation, suggesting inhibition of translation as a common pathogenic mechanism. Future research aimed at elucidating the molecular mechanisms underlying the translation defect induced by CMT‐mutant aaRSs should provide novel insight into the molecular pathogenesis of these incurable diseases.


Introduction: Aminoacyl tRNA synthetases catalyze the first step of protein synthesis
Protein translation involves the matching of triplet codons in the mRNA with anticodons of tRNAs. This job is performed by the ribosome, which subsequently transfers the nascent peptide chain to the amino acid attached to a matching tRNA. Therefore, the accuracy of the amino acid sequence of a protein depends on three factors: the flawlessness of the mRNA coding sequence, the correct matching of codon and anticodon by the ribosome, and the correct attachment of amino acids to the tRNA. Aminoacyl tRNA synthetases (aaRSs) are the enzymes that catalyze the covalent attachment of amino acids to their cognate tRNAs in a two-step reaction (Fig. 1A) [1,2]. After synthesis, aminoacyl-tRNAs are delivered to the ribosome by elongation factors. It is thought that during the translation cycle, tRNAs are always chaperoned by aaRSs, elongation factors, or other proteins that directly interact with aaRSs, and never freely diffuse in the cytoplasm of mammalian cells [3].
Cytoplasmic aaRSs constitute a family of 19 enzymes, one for each amino acid, with the exception of the bifunctional glutaminyl-prolyl-tRNA synthetase. Mitochondrial aaRSs are usually encoded by separate genes, except for GlyRS and LysRS, for which the cytoplasmic and the mitochondrial enzymes are encoded by a single gene and generated by the use of alternative translational start sites or by alternative mRNA splicing [4][5][6]. aaRSs are grouped into two structural classes (I and II) based on the architecture of the catalytic core domains (Table 1) [7,8]. Class I aaRSs are monomeric or homodimeric, and TyrRS and TrpRS function as obligate homodimers, in which the tRNA binds across the subunits [9] (Table 1). Most class II aaRSs are homodimeric. Each aaRS class is further subdivided into subclasses (Table 1) [10,11].
All aaRSs contain a catalytic and an anticodon binding domain, which contacts the anticodon of the cognate tRNA. A number of aaRSs further contain dimerization or multimerization domains, editing domains, which mediate editing of mischarged tRNAs (Box 1), protein interaction domains, and domains responsible for subcellular localization. Apart from their canonical tRNA aminoacylation function, several aaRSs acquired additional functional domains during evolution, which mediate noncanonical functions in a diversity of biological processes, including inflammation, transcriptional regulation, translational regulation, apoptosis, rRNA transcription, angiogenesis, cellsignaling, autoimmune response, tRNA maturation, and mitochondrial RNA splicing [10,12,13]. Finally, nine cytoplasmic aaRSs and three auxiliary proteins (aaRS-interacting multi-functional proteins or AIMPs) associate to form the "multi-synthetase complex" (MSC) ( Table 1), which may facilitate protein translation. In addition, some MSC components are released from the MSC upon specific signals, whereupon they exert non-canonical activities [10,14].

Dominant mutations in tRNA synthetase genes cause Charcot-Marie-Tooth peripheral neuropathy
Over the past 12 years, heterozygous mutations in five distinct tRNA synthetase genes have been identified as a genetic cause of Charcot-Marie-Tooth (CMT) peripheral neuropathy. This genetic disorder is characterized by selective degeneration of peripheral motor and sensory axons, leading to progressive muscle weakness and wasting, and decreased sensation of vibration, touch, and pain. CMT patients typically display reduced or absent deep-tendon reflexes and foot deformities such as high arches (pes cavus) and hammer toes. Disease onset is variable, but usually occurs in the first two decades of life and subsequently shows a slow progression over decades, patients ultimately becoming wheelchair-bound. Typically, distal muscles are affected first, with a disto-proximal progression over time [15]. With a prevalence of one in 2,500 individuals, CMT is the most common inherited neuromuscular disorder [16].
Traditionally, a distinction is made between demyelinating forms of CMT Figure 1. Mutations in tRNA synthetases cause CMT peripheral neuropathy. A: aaRSs catalyze tRNA aminoacylation in a two-step reaction. In the first step (1), the amino acid (AA) is activated with ATP, resulting in the formation of an enzyme-bound aminoacyl-adenylate (AA-AMP), with the concomitant release of pyrophosphate (PPi). At this stage, aaRSs with pre-transfer editing activity can remove misacetylated aminoacyl adenylate. In the second step (2), the aminoacyl-adenylate is transferred to the tRNA, with release of AMP. aaRSs which posses posttransfer editing activity can hydrolyze misacetylated tRNAs to prevent amino acid misincorporation in nascent proteins. B: Distribution of CMT-associated mutations in aaRSs. Schematic representation of GlyRS, TyrRS, AlaRS, HisRS, and MetRS proteins and their functional domains. CMT-associated mutations found in single patients are indicated in black, while mutations that co-segregate with disease in CMT families are labeled red. Mutations labeled green are equivalent to mutations in mouse CMT2D models. For GlyRS, the positions of the mutations refer to the cytoplasmic form of the human protein. (CMT1) and axonal forms (CMT2). Clinically, this distinction is based on the measurement of median or ulnar nerve conduction velocities (NCVs): severely reduced NCVs of <38 m/s are classified as CMT1, whereas normal or mildly reduced NCVs (>38 m/s) are classified as CMT2 [17]. CMT1 is pathologically characterized by segmental demyelination and remyelination with formation of so-called onion bulbs: concentric arrangements of supernumerary Schwann cells around an incompletely remyelinated axon [18]. The primary demyelination in CMT1 ultimately leads to axonal degeneration, giving rise to the classical CMT symptoms [19]. CMT1 is the most prevalent, while CMT2 accounts for $20% of cases [20]. CMT2 is electrophysiologically characterized by reduced compound action potential amplitudes, and pathologically by chronic axonal degeneration and regeneration [19]. More recently, the existence of intermediate forms of CMT has been recognized, characterized by intermediate NCVs (25-45 m/s), and pathological features of both demyelination and axonal degeneration [21]. CMT is not only clinically, but also genetically heterogeneous: mutations in more than 70 genes have been identified as causative for CMT [22]. aaRSs were first implicated in CMT in 2003, when four heterozygous mutations in GARS (E71G, L129P, G240R, G526R), encoding cytoplasmic GlyRS, were reported as causative for CMT type 2D (CMT2D) and distal hereditary motor neuropathy type Va (HMN5A) [23]. HMN5A is phenotypically similar to axonal CMT, but without sensory involvement [24]. A series of follow-up studies reported additional mutations  The accuracy of tRNA aminoacylation is essential for correct translation of mRNA triplet codons into the primary amino acid sequence of proteins. This requires correct discrimination of both the amino acid and the tRNA by the aaRS. Apart from recognition of the tRNA anticodon by the anticodon binding domain of the aaRS, several additional interactions between aaRS and tRNA typically occur, enabling correct recognition of the cognate tRNA [2,10]. It can be much more challenging for aaRSs to discriminate between cognate and structurally similar, non-cognate amino acids [10]. Therefore, 10 of 19 cytoplasmic aaRSs also possess proofreading or editing activity (Table 1). Two modes of editing can be distinguished: (i) pre-transfer editing, which removes misactivated aminoacyl adenylate that is produced in the first step of the aminoacylation reaction; and (ii) post-transfer editing, which involves hydrolysis of mis-aminoacylated tRNAs (Fig. 1A) [10,11,98]. Post-transfer editing requires the presence of a spatially separate editing domain, which is present in 7 of 10 editing aaRSs. SerRS, MetRS, and LysRS lack such a domain and catalyze pre-transfer editing only ( Table 1). The discovery of editing activity by aaRSs has led to the "double-sieve" model to explain the accuracy of tRNA aminoacylation: the active site serves as the first sieve, activating cognate, isosteric, and smaller amino acids but excluding larger amino acids, and the editing site is the second sieve, hydrolyzing misactivated or mischarged amino acids but rejecting cognate amino acids [10,98].
Next, heterozygous missense mutations in AARS, encoding cytoplasmic AlaRS, were reported as causative for CMT type 2N (CMT2N). A R329H mutation segregated with disease in two unrelated families [36]. Five additional mutations were subsequently identified in patients with axonal CMT (N71Y and G102R) [37,38], intermediate CMT (E688G) [39], distal hereditary motor neuropathy (D893N) [40], and a family affected by rippling muscles and cramps that included one member that also exhibited axonal CMT (E778A) [41]. These six mutations are distributed throughout the AlaRS primary sequence (Fig. 1B). Importantly, five of the six mutations were shown to segregate with disease.
Finally, mutations in HARS and MARS were associated with CMT. Four heterozygous HARS mutations, all mapping to the catalytic domain of HisRS, segregated with peripheral neuropathy in four unrelated families: T132I, P134H, D175G, and D364Y [42] (Fig. 1B). The associated phenotypic spectrum was broad, encompassing axonal CMT, hereditary motor neuropathy and intermediate CMT. Several additional missense variants in HARS have been identified in peripheral neuropathy patients, but their pathogenicity is unclear [43]. In a family with late-onset CMT2, two affected family members were heterozygous for a R618C mutation in MARS, but the 85 year-old mother of the index patient also carried the mutation but was unaffected. Thus, this mutation is either not disease-causing, or displays incomplete penetrance [44]. In addition, a single patient from a late-onset CMT family and an additional patient from an early-onset CMT family carried a heterozygous P800T mutation in MARS [35,45]. Hence, it is uncertain if these MARS variants are pathogenic, as segregation with disease was not reported (Fig. 1B).
Interestingly, whereas heterozygous mutations in cytoplasmic aaRS genes are consistently associated with axonal CMT and its phenotypic variants, homozygous or transheterozygous mutations in these genes typically induce more severe syndromes, often involving multiple organ systems ( Table 2). In some cases, peripheral neuropathy can be a component of these severe phenotypes. Lastly, recessive mutations in several mitochondrial aaRS genes give rise to a variety of disorders, which were recently reviewed [13].

Animal models for CMT associated with aaRS mutations
A number of animal models for CMT associated with aaRS mutations (CMT-aaRS) have been generated, which recapitulate several characteristics of the human disease. Two mouse CMT2D models originated from independent ENU mutagenesis screens [46,47]. In the first model, mice heterozygous for a substitution of Pro278 by Lys and Tyr (P278KY) in GlyRS (corresponding to P234KY in human cytoplasmic GlyRS) exhibited overt neuromuscular dysfunction by three weeks of age and a greatly shortened life span [46]. Gars P278KY/+ mice displayed abnormal neuromuscular junction (NMJ) morphology, more pronounced in distal muscles. Neuromuscular transmission was impaired, and nerve conduction velocities reduced. Loss of large diameter peripheral motor and sensory axons was evident, more pronounced distally and without myelination defects or spinal cord pathology [46]. In the second model, mice heterozygous for a GlyRS C201R mutation (corresponding to C157R in human cytoplasmic GlyRS) showed loss of grip strength, diminished fine motor control, and reduced hindlimb muscle force, with an apparently normal life span [47]. Muscle atrophy and NMJ morphology defects were evident, and the diameter of peripheral sensory axons was reduced, leading to reduced NCVs [47]. Interestingly, NMJ denervation in Gars P278KY/+ and Gars C201R/+ mice is preceded by NMJ maturation defects [48].
Apart from CMT2D mouse models, a number of Drosophila CMT-aaRS models have been reported, which are ideally suited to conduct genetic screens for putative disease-modifying genes. A Drosophila DI-CMTC model was generated by expression of human TyrRS using the UAS/GAL4 system, which allows for spatial and temporal control of transgene expression [49]. In contrast to wild type (WT) TyrRS, expression of CMT-mutant TyrRS (G41R, 153-156delVKQV, and E196K) induced progressive motor deficits, electrophysiological evidence of neuronal dysfunction, and axonal degeneration. Not only ubiquitous, but also neuron-selective expression of mutant TyrRS induced these phenotypes, indicating that the mutant enzymes are intrinsically toxic to neurons [50].
More recently, several independent Drosophila CMT2D models have been reported. One set of models involves transgenic overexpression of the cytoplasmic form of Drosophila GlyRS with a C-terminal HA tag, either WT or with G240R or P234KY mutations. Neuronselective expression of GlyRS_G240R induced an age-dependent, mild climbing defect, and selective expression of mutant GlyRS transgenes in the "giant fiber" system [51] triggered electrophysiological and morphological defects in the giant fiber axon terminal [52]. In larvae, mutant Drosophila GlyRS reduced the frequency of larval body wall muscle contractions when expressed either in muscle, in neurons, or ubiquitously [53]. The second set of Drosophila CMT2D models utilizes transgenes that allow expression of both the cytoplasmic and mitochondrial forms of human GlyRS, carrying E71G, G240R, G526R, or no mutations [54]. Ubiquitous expression of mutant, but not WT, transgenes from the adult stage onwards greatly reduced life span, and motor neuron selective expression impeded climbing behavior and induced NMJ morphology defects and progressive muscle denervation, distal muscles being more severely affected. Selective expression of mutant GlyRS in sensory neurons induced morphology defects [54]. Apart from the expression of mixed cytoplasmic and mitochondrial forms of human GlyRS versus cytoplasmic Drosophila GlyRS, the two sets of models distinguish themselves by the fact that the human GlyRS transgenes are untagged, and introduced into specific genomic landing sites, which uniformizes transgene expression levels [55,56]. Taken together, the available Drosophila and mouse models form complementary tools to study the molecular pathogenesis of CMT-aaRS and they have significantly contributed to our current understanding of disease pathogenesis.
How could mutant aaRSs cause peripheral neuropathy?
Could partial loss of aminoacylation activity underlie CMT-aaRS?
It was initially hypothesized that CMTcausing aaRS mutations may lead to loss of aminoacylation activity. Since patients are heterozygous for CMT-aaRS mutations, this could lead to a reduction of "overall" aminoacylation activity, either through haploinsufficiency or a dominant negative mechanism. This may deplete the pool of aminoacylated cognate tRNAs, so that, when below a critical threshold, the supply of this tRNA species to the ribosome would become insufficient, leading to ribosome stalling at codons for the cognate amino acids, thus inhibiting translation. This is a realistic scenario, as recently shown by a mouse mutant, in which diminished amounts of a brain-specific Arg-tRNA Arg causes ribosome stalling at Arg codons, which is exacerbated by the absence of Gtpbp2, a protein functioning to resolve stalled ribosomes, leading to severe neurodegeneration [57]. This hypothesis was further supported by the observation that almost all of the amino acid residues mutated in CMT-aaRS are highly conserved during evolution: 31 of the 34 mutated residues are conserved at least as far as Drosophila melanogaster (Table 3). This leads to the stunning observation that in Drosophila GlyRS 16 of the 17 CMTassociated residues are strictly conserved, whereas the overall amino acid identity is only 60%. For TyrRS and HisRS, all disease-associated residues are at least conserved to yeast (Table 3). This suggests that interference with an ancient, important or even essential function of these enzymes, most probably aminoacylation, may underlie CMT pathogenesis.
It is possible that for some CMT-aaRS mutations, partial loss of aminoacylation activity may cause or causally contribute to peripheral neuropathy phenotypes. However, at least for some CMT-aaRS mutations, several lines of evidence have shown that loss of aminoacylation activity is not required to cause CMT. First, direct analysis of aminoacylation activity, either using in vitro aminoacylation assays or in vivo genetic complementation assays in yeast or Drosophila, revealed that several CMT-aaRS mutations result in loss or severe reduction of aminoacylation activity, but some mutations, which segregate with disease in families, do not affect aminoacylation activity, including GlyRS E71G, TyrRS E196K, and AlaRS E778A [41,50,54,[58][59][60] (Table 3). Furthermore, in CMT2D mouse models, heterozygous P278KY and C201R mutations in GlyRS do not reduce tRNA Gly aminoacylation activity [46,47,59]. Secondly, if reduction of aminoacylation activity would underlie CMT pathogenesis, transgenic increase of WT aaRS expression should rescue peripheral neuropathy in CMT-aaRS animal models. This was not the case in CMT2D mouse models [61]. Thirdly, in case of a haploinsufficient mechanism, animals heterozygous for aaRS loss-of-function alleles should develop peripheral neuropathy. However, heterozygosity for a Gars loss-offunction allele in mice or a TyrRS null allele in flies did not induce peripheral neuropathy phenotypes [46,50]. Finally, overexpression of mutant human GlyRS in Drosophila induced peripheral neuropathy phenotypes, without reduction of tRNA Gly aminoacylation activity and without altering the in vivo ratio of aminoacylated versus non-aminoacylated tRNA Gly [54]. Taken together, this leaves us with two possible scenarios: (i) all CMT-aaRS mutations result in the acquisition of a novel, toxic property that underlies peripheral neuropathy; or (ii) some CMT-aaRS mutations cause CMT through a gain-of-toxic-function mechanism, whereas other CMT-aaRS mutations cause CMT through partial loss of aminoacylation activity, most likely through a dominant-negative mechanism. Further research is needed to distinguish between these two scenarios.
tRNA misacylation leading to misincorporation of amino acids in proteins is unlikely to underlie CMT-aaRS A second possible mechanism is that CMT-aaRS mutations could lead to an increased frequency of tRNA misacylation, either by reducing the ability of aaRSs to discriminate cognate from non-cognate amino acids, or by impairing the pre-or post-transfer editing activity. tRNA misacylation would lead to misincorporation of amino acids in proteins, leading to protein misfolding and aggregation. The plausibility of this mechanism is illustrated by the mouse "sticky" mutant, in which a A734E mutation in the AlaRS editing domain compromises the proofreading activity of this enzyme, resulting in cerebellar Purkinje cell loss and ataxia, and intracellular accumulation of misfolded, ubiquitinated proteins in neurons [62]. Similarly, in Drosophila, a double mutation in PheRS, which both impairs the capacity to discriminate Phe from Tyr and disrupts the postediting activity, leads to misacylation of tRNA Phe with Tyr, resulting in protein mistranslation and ER stress. The mutant flies exhibit several defects, including neuronal loss, impaired locomotor performance, shorter life span, and smaller organ size [63].
However, there are several arguments against this hypothesis. Firstly, some CMT-aaRS mutations disrupt the binding site for amino acids or ATP, resulting in an aaRS no longer able to activate amino acids, cognate or noncognate. This was shown for GlyRS G526R [64] and TyrRS G41R and 153-156delVKQV mutations [60]. With respect to defective editing, of the five CMT-associated aaRSs, only AlaRS has post-editing activity, while MetRS and AlaRS have pre-editing activity (Table 1). CMT mutations in these editing aaRSs have not been reported to affect the editing process. In particular, the AlaRS_E778A mutation affects neither aminoacylation nor editing activity [41]. Finally, sticky mice exhibit cerebellar ataxia but no peripheral neuropathy. Vice versa, Gars P278KY/+ mice do not show cerebellar neurodegeneration, and there is no evidence for accumulation of misfolded proteins in these mice [65]. Thus, it seems unlikely that translational infidelity due to tRNA misacylation underlies CMT-aaRS.

Alteration of aaRS dimerization is unlikely to cause CMT-aaRS
Interestingly, when mapped on the GlyRS crystal structure, CMT2D-causing mutations cluster around the dimer interface, suggesting that alteration of GlyRS dimer formation may be involved in disease pathogenesis [59]. However, different CMT-GlyRS mutations have different effects on dimerization: some mutations strengthen dimer formation, others prevent dimer formation, and yet others do not influence dimerization [59,64,66]. Therefore, it is unlikely that altered dimer formation causes CMT2D.

Could mislocalization of mutant aaRSs contribute to CMT-aaRS?
The fact that loss of aminoacylation activity is not required to cause CMT does not exclude the possibility that CMT-associated mutations could result in subcellular mislocalization of aaRSs, possibly causing defects in local protein translation. For instance, reduced localization of CMT-mutant aaRSs to axons and/or nerve endings could lead to impaired local protein synthesis and axonal degeneration. This is a conceivable scenario, as endogenous GlyRS and TyrRS proteins are localized to motor neuron cell bodies, axons and nerve endings in human and mouse nervous tissues, and in cultured mouse motor neurons [34,58,65]. Consistently, a number of CMT-mutant GlyRS and TyrRS proteins displayed altered subcellular distribution in neuronal cell lines, which can be differentiated to form neurite projections. Specifically, in the mouse motor neuron, neuroblastoma fusion cell line MN-1, some CMT GlyRS mutants exhibited altered subcellular localization, including L129P, S211F, P234KY, G240R, P244L, I280F, H418R, and G598A GlyRS [58,65,67]. However, other GlyRS mutants (A57V, E71G, D146N, D500N, and G526R) showed a similar subcellular distribution as WT GlyRS [58,67].
Different from studies in cell lines, the subcellular localization of mutant GlyRS and TyrRS in CMT animal models was generally reported to be indistinguishable from the respective WT proteins. In CMT2D mouse models, GlyRS localization in sciatic nerve fibers and in spinal cord sections was unaltered [65]. However, in vivo expression of WT, L129P, or G240R human GlyRS in mouse motor and sensory neurons by viral gene transfer revealed WT GlyRS localization along the length of sciatic nerve axons, whereas localization of L129P and G240R GlyRS to sciatic nerve axons was diminished [68,69]. In Drosophila CMT2D models, HA-tagged WT and P234KY Drosophila GlyRS showed similar subcellular localization in giant fiber axons and motor neuron cell bodies and axons [52,53], and WT, E71G, G240R, and G526R human GlyRS displayed similar subcellular distribution in motor neurons, with diffuse localization to motor neuron cell bodies, axons, and NMJs [54]. Furthermore, in Drosophila DI-CMTC models, WT, G41R, 153-156delVKQV, and E196K human TyrRS displayed similar distribution in both motor and sensory neurons, with homogeneous distribution throughout the cell body, axon, and major dendrite branches [54].
The discrepancy between in vitro and in vivo studies may partly be due to effects of protein tagging and overexpression, as well as the fact that aaRS distribution may differ between differentiating neuronal cells in vitro and mature motor and sensory neurons in vivo. For instance, the GlyRS P234KY mutant mislocalizes in MN-1 and N2a cells, but displays normal subcellular localization in the mouse spinal cord [59,65]. Importantly, even in neuronal cell lines, some CMT-mutant aaRSs display normal subcellular localization, including GlyRS A57V, E71G, D146N, and AlaRS E778A. Thus, at least for these mutants, defects in local protein translation due to subcellular mislocalization are unlikely. Furthermore, the peripheral neuropathy phenotypes in mouse and Drosophila CMT-aaRS models are not attributable to subcellular mislocalization of mutant aaRSs. Further studies are needed to evaluate whether for some mutant aaRSs, subcellular mislocalization could contribute to CMT pathogenesis. In this respect, it would be highly interesting to investigate the subcellular localization of mutant aaRSs in CMT-aaRS patient tissues.
Could interference with noncanonical functions of aaRSs cause peripheral neuropathy?
As indicated above, several aaRS have acquired additional functions during evolution, often through incorporation of additional functional domains. For the CMT-associated aaRSs, functions beyond aminoacylation have been described for GlyRS, TyrRS, MetRS, and HisRS [70][71][72][73][74][75][76][77], and CMT-causing mutations could possibly interfere with these functions. The fact that missense mutations in aaRSs can affect non-canonical functions is illustrated by a Y341A mutation in TyrRS, which uncovers an internal ELR tripeptide, thereby activating the cytokine function in the fulllength TyrRS protein, which is normally inactive as a cytokine [78]. Unfortunately, the effect of CMT-causing mutations on GlyRS, TyrRS, MetRS, and HisRS non-canonical functions has thus far not been investigated. Although this mechanism might contribute to CMT pathogenesis for some mutations, it seems unlikely that it represents a common pathogenic mechanism underlying CMT-aaRS, for a number of reasons. First, for AlaRS, non-canonical functions have not been reported. Secondly, different aaRSs acquired distinct non-canonical functions. Thus, if alteration of non-canonical functions would underlie CMT-aaRS, this would imply that different mutant aaRS would cause CMT through distinct molecular mechanisms. Although not impossible, this seems unlikely. Furthermore, for GlyRS, CMT mutations are distributed throughout the protein, making it unlikely that all mutations affect GlyRS non-canonical functions. Thus, to date, no concrete data suggest that alteration of non-canonical aaRS functions contribute to CMT pathogenesis.

A gain-of-toxic-function mechanism likely underlies CMT-aaRS
In CMT2D mouse models, convincing genetic evidence indicates that mutant GlyRS proteins cause peripheral neuropathy by a "toxic-gain-of-function" mechanism. Genetically, the characteristics of a toxic-gain-of-function (neomorphic) allele are that phenotypes are not modified by altering the levels of WT protein, but enhanced by increasing the levels of mutant protein. Consistently, transgenic overexpression of WT GlyRS does not improve the neuropathy phenotype in heterozygous Gars P278KY/+ and Gars C201R/+ mice [61]. Furthermore, homozygous Gars C201R/C201R and transheterozygous Gars C201R/P278KY mice in a WT GlyRS overexpression background display enhanced peripheral neuropathy phenotypes [61]. Similarly, in Drosophila DI-CMTC and CMT2D models, the severity of peripheral neuropathy phenotypes is transgene dosage-dependent [50,53,54]. Molecularly, a toxicgain-of-function mechanism can involve novel protein-protein interactions enabled by the disease-causing mutations, in which the WT protein does not engage. These novel protein-protein interactions could affect the function of the interacting protein(s), thereby causing disease. Interestingly, several spatially dispersed GlyRS mutations (L129P, G240R, G526R, and G598A) induce the same conformational opening of a consensus area that is mostly buried in WT GlyRS [66]. A possible molecular mechanism underlying the toxic-gain-of-function of CMTmutant GlyRS was recently reported [79], as several CMT-GlyRS mutants, including E71G, L129P, P234KY, and G240R, strongly bound to neuropilin-1 (Nrp1), a co-receptor for both semaphorins and VEGF-A [80]. In contrast, WT GlyRS only weakly bound to Nrp1. VEGF-A was previously implicated in motor neuron degeneration, as low VEGF-A expression leads to adult-onset motor neuron degeneration in mice, reminiscent of human amyotrophic lateral sclerosis (ALS) [81], and exogenous VEGF-A administration has significant therapeutic effects in ALS rodent models [82,83]. CMT-GlyRS mutants competed with VEGF-A for binding to Nrp1, and heterozygosity for Nrp1 enhanced the peripheral neuropathy phenotype of Gars P278KY/+ mice. Furthermore, increasing VEGF-A expression in hindlimb muscles improved motor performance of Gars P278KY/+ mice [79]. Although it remains to be investigated whether all CMT-causing mutations increase GlyRS binding to Nrp1, this mechanism illustrates how CMTmutant, misfolded GlyRS may interfere with signaling pathways that are critical for survival of peripheral motor and sensory axons. It is likely that other CMT-mutant aaRSs harbor similar neomorphic activities, and unraveling their molecular mechanisms is a major challenge that the CMT-aaRS field is currently facing. Of note, the currently available evidence does not exclude the possibility that some CMT-aaRS mutant  proteins might cause peripheral neuropathy through partial loss of aminoacylation activity.
Inhibition of protein translation independent of aminoacylation: A common pathogenic mechanism of CMT-aaRS?
The toxic-gain-of-function mechanism may involve aberrant interactions with components of the protein translation pathway or, alternatively, with pathways unrelated to translation. The hypothesis that mutant aaRSs could affect translation was recently investigated in Drosophila CMT2D and DI-CMTC models, utilizing a novel method based on noncanonical amino acid tagging, which allows to cell-type-specifically monitor translation in vivo ( Fig. 2A) [54,84]. Remarkably, selective expression of three distinct GlyRS mutants (E71G, G240R, G526R) in either motor or sensory neurons dramatically inhibited global protein translation [54]. 35 S-methionine incorporation confirmed impaired translation when mutant GlyRS was ubiquitously expressed in adult flies. Furthermore, selective expression of three distinct CMT-TyrRS mutants (G41R, del153-156VKQV, E196K) also significantly inhibited translation in motor and sensory neurons, suggesting that impaired translation may constitute a common pathogenic mechanism underlying CMT-aaRS [54].
Consistent with a toxic-gain-of-function mechanism, inhibition of translation by mutant GlyRS was independent of tRNA Gly aminoacylation. Indeed, expression of mutant GlyRS transgenes -in an otherwise WT Drosophila GlyRS background -did not reduce overall aminoacylation activity and did not alter the in vivo ratio of glycylated versus non-glycylated tRNA Gly , with >90% Gly-tRNA Gly . Moreover, co-overexpression of WT Drosophila GlyRS did not rescue the translation defect induced by mutant human GlyRS [54]. Finally, it is likely that impaired translation causally contributes to peripheral neuropathy phenotypes in Drosophila CMT-aaRS models, as inhibition of translation independent of mutant aaRS expression -by expression of constitutively active forms of eukaryotic initiation factor 4E binding protein (4E-BP) -was sufficient to induce muscle denervation and sensory neuron morphology defects [54]. Thus, mutant aaRSs inhibit translation by a toxicgain-of-function mechanism, independent of aminoacylation, and impaired translation may causally contribute to peripheral neuropathy phenotypes (Fig. 2B). It will now be important to evaluate whether translation is also affected in mouse CMT2D models, and to identify the molecular mechanism by which mutant aaRSs inhibit translation (Box 2).
Interestingly, regulation of translation by aaRSs through non-catalytic mechanisms have been previously reported. Some aaRSs regulate the translation of their own transcript, others regulate translation of a select number of transcripts different from their own, and phosphorylation of MetRS provides a mechanism for global regulation of translation [85]. For instance, E. coli ThrRS binds as a homodimer to two stem-loop structures in the 5 0 UTR of its own mRNA, which mimic the anticodon arm of tRNA Thr [86,87], thereby preventing ribosome binding and inhibiting translation initiation [88]. As tRNA Thr and ThrRS mRNA compete for binding to ThrRS, this constitutes a negative autoregulatory mechanism [88][89][90]. Negative autoregulation of translation was also reported for yeast AspRS, which binds to the 5 0 UTR of its transcript that adopts a tRNA Asp anticodon-like structure [91,92].
Gene-specific regulation of translation by aaRSs is exemplified by GluProRS, which is a component of the interferongamma-activated inhibitor of translation (GAIT) complex. This heterotetrameric complex suppresses translation of selected mRNAs in interferon-gammaactivated monocytic cells [93]. In response to interferon-gamma, GluProRS is phosphorylated and released from the MSC and incorporated in the GAIT

Box 2
Proposed experimental approaches to verify the hypothesis The central hypothesis proposed here is that inhibition of protein translation is a common pathogenic mechanism underlying CMT-aaRS. In order to test this hypothesis, several aspects would need experimental verification: (1) To determine whether all CMT-aaRS mutant proteins inhibit translation in vivo, it would be necessary to generate additional CMT-aaRS animal models, including CMT-AlaRS and CMT-HisRS models. NCAT technology could then be used to evaluate whether protein translation is inhibited in motor and sensory neurons of these models. (2) To study the relevance of the findings in Drosophila models for human CMT-aaRS, it should be evaluated whether protein translation is also affected in CMT-aaRS mouse models, and/or in induced pluripotent stem cell (iPSC)-derived motor and sensory neurons from CMT-aaRS patients. (3) A first step to gain insight into the molecular mechanism underlying the translation defect in CMT-aaRS Drosophila models could be to determine whether CMT-mutant aaRSs interfere with upstream regulatory pathways of translation, or rather directly with translation initiation or elongation. This could be done by genetic manipulation of known key regulators of these processes in Drosophila CMT-aaRS models, and evaluating the effect on protein translation by NCAT.
More broadly, it will be important to determine whether all CMT-mutant aaRSs cause disease through a gain-of-toxic-function mechanism, or, alternatively, whether some mutant aaRSs cause disease through loss-offunction and some through gain-of-toxic-function. This could be done by generating additional CMT-aaRS animal models and testing whether overexpression of the relevant wild type aaRS rescues the peripheral neuropathy phenotypes.

E. Storkebaum
Insights & Perspectives ..... complex, in which it is the subunit that binds to the 3 0 UTR of target mRNAs, resulting in translational silencing of target mRNAs [94]. In this case, tRNA mimicry is not involved in target mRNA binding, as the WHEP domains in GluProRS are responsible for binding of the GAIT element stem loop [95]. Interestingly, an additional level of translational regulation of GAIT target genes involves the production of a truncated form of GluProRS, which shields GAIT-element bearing transcripts from the GAIT complex, thereby countering translational repression [96]. The fact that aaRSs can not only inhibit, but also activate translation is illustrated by the binding of GlyRS to the poliovirus IRES, which promotes the accommodation of the ribosome and greatly enhances IRES activity. Poliovirus IRES uses tRNA Gly anticodon stem-loop mimicry to recruit GlyRS [97].
Finally, phosphorylation of human MetRS on Ser 662 by GCN2 reduces its catalytic activity due to diminished tRNA i Met binding, leading to downregulation of global translation [73]. Since expression of CMT-mutant GlyRS and TyrRS in Drosophila motor and sensory neurons inhibited global rather than gene-specific translation, only the latter mechanism may possibly mediate translational inhibition in Drosophila CMT models. However, since the in vivo ratio of aminoacylated versus non-aminoacylated tRNA i Met was unaltered in larvae that ubiquitously expressed mutant GlyRS [54], it seems unlikely that this mechanism is involved.

Conclusions and prospects
Over the past 12 years, dominant mutations in five distinct aaRS genes have been associated with CMT peripheral neuropathy, and significant progress has been made toward understanding how mutations in these ubiquitously expressed, essential enzymes may lead to selective degeneration of peripheral motor and sensory axons. It became evident that some CMT-causing aaRS mutations do not affect aminoacylation activity, showing that loss of aminoacylation activity is not required to cause peripheral neuropathy. Furthermore, mislocalization of CMT-mutant aaRSs has been reported in cultured neuronal cell lines, but not in CMT-aaRS animal models, and is, therefore, not necessary to induce CMT phenotypes. It is further unlikely that misincorporation of amino acids in proteins due to tRNA misacylation contributes to CMT-aaRS pathogenesis. Rather, convincing genetic evidence in CMT-aaRS animal models has shown that a gain-of-toxicfunction mechanism underlies CMT-aaRS pathogenesis, and interference with VEGF-Nrp1 signaling is a possible molecular mechanism contributing to CMT-GlyRS. Moreover, impaired translation may be a common pathogenic event in CMT-aaRS, as all of six CMTmutant GlyRS and TyrRS proteins inhibited translation in Drosophila motor and sensory neurons. This translational slowdown was independent of tRNA aminoacylation and caused by a gain-of-toxic-function mechanism. The molecular mechanism by which mutant aaRSs inhibit translation should be the focus of future research (Box 2). It is further important to confirm that translational defects are also present in CMT-aaRS mouse models (Box 2). Finally, another outstanding question is to which extent distinct mutations in distinct aaRSs cause peripheral neuropathy through common molecular mechanisms (Box 2). For instance, it remains possible that for some mutations, loss of aminoacylation activity may contribute to axonal degeneration. Overall, if future research could provide detailed molecular insights into CMT-aaRS pathogenesis, this may form the first step toward the development of an effective drug treatment for this incurable disorder.