Aminoacyl‐tRNA synthetases in Charcot–Marie–Tooth disease: A gain or a loss?

Abstract Charcot‐Marie‐Tooth disease (CMT) is one of the most common inherited neurodegenerative disorders with an increasing number of CMT‐associated variants identified as causative factors, however, there has been no effective therapy for CMT to date. Aminoacyl‐tRNA synthetases (aaRS) are essential enzymes in translation by charging amino acids onto their cognate tRNAs during protein synthesis. Dominant monoallelic variants of aaRSs have been largely implicated in CMT. Some aaRSs variants affect enzymatic activity, demonstrating a loss‐of‐function property. In contrast, loss of aminoacylation activity is neither necessary nor sufficient for some aaRSs variants to cause CMT. Instead, accumulating evidence from CMT patient samples, animal genetic studies or protein conformational analysis has pinpointed toxic gain‐of‐function of aaRSs variants in CMT, suggesting complicated mechanisms underlying the pathogenesis of CMT. In this review, we summarize the latest advances in studies on CMT‐linked aaRSs, with a particular focus on their functions. The current challenges, future direction and the promising candidates for potential treatment of CMT are also discussed.


| INTRODUC TI ON
Charcot-Marie-Tooth disease (CMT) is one of the most common inherited neuromuscular disorders with an estimated prevalence of 1/2,500 worldwide (Rossor et al., 2016;Skre, 1974). The major clinical manifestation of CMT is the degeneration of both motor and sensory peripheral nerves, leading to a loss of muscle tissue and touch sensation in bodily extremities (Patzko & Shy, 2011).
Based on electrophysiological and histopathological criteria, CMT is divided into two major groups: demyelinating CMT type 1 (CMT1) and axonal CMT type 2 (CMT2). CMT1 is the most prevalent type presented as demyelinating peripheral neuropathy, which manifests as markedly decreased nerve-conduction velocities. In contrast, CMT2 accounts for approximately 20% of cases characterized by pathological axonal loss at nerve biopsy and has relatively normal conduction velocities (Pareyson & Marchesi, 2009) In typical cases, CMT onset often occurs in the first two decades of life, and the disease processes slowly without affecting life expectancy (Laura et al., 2019;Pareyson & Marchesi, 2009). However, the age of onset, disease course and severity vary greatly based on the CMT subtype, causative genes and types of variants. Despite significant advances in the genetic diagnosis of CMT using next-generation sequencing technology, no effective therapies have thus far been developed.
The aaRS family comprises ubiquitously expressed enzymes that are involved in the translation of the genetic code by charging amino acids onto their cognate tRNAs during protein synthesis (Ling et al., 2009). Based on structural features, 20 canonical aaRSs are divided into two classes that differ in the architecture of their active sites for adenylate synthesis. Class I aaRSs contain a typical Rossmann-fold utilized for nucleotide binding and two well-conserved signature sequences (HIGH and KMSKS), whereas class II aaRSs are less conserved and include three signature motifs (Motifs 1-3) within a seven-stranded β-sheet and three flanking α-helices.
Among the five known CMT-associated aaRSs, TryRS and TrpRS belong to class I, and the other three (GlyRS, AlaRS and HisRS) belong to class II.
Intriguingly, variants of these tRNA synthetase genes are strongly associated with CMT, but not all variants affect the aminoacylation activities of the enzymes. More strikingly, CMT-like neuropathy phenotypes in animal models generated by some variants cannot be rescued by overexpression of wild type (WT) proteins, suggesting that CMT is very likely linked to toxic gain-of-function associated with the variants themselves. In this review, we discuss the latest advances in studies on aaRSs in CMT, particularly focusing on their roles in the pathogenesis of the disease.

| GARS VARIANTS
GlyRS is a class II aaRS with a conserved catalytic domain composed of a central antiparallel β-sheet flanked with α-helices and three conserved sequence motifs , followed by an anticodon domain. The N-terminal extension contains a specific appended helix-turn-helix structural motif named the WHEP domain, which derives from three of the five WHEP-containing proteins TrpRS, HisRS and EPRS   (Figure 1a).
GlyRS was the first tRNA synthetase implicated in CMT (Antonellis et al., 2003). To date, more than twenty missense variants of GARS have been discovered in patients with CMT2D (OMIM# 601472), an autosomal-dominant axonal subtype of CMT, or dHMN type V (dHMN-V, OMIM# 600794), a subtype of dHMN with upperlimb predominance (Motley et al., 2010) (Table 1) and GARS G598A ) are closely associated with the disease; however, some affect aminoacylation activity, while others do not (Griffin et al., 2014) (Table 1). For example, GARS E71G segregates with CMT2D in large families but has WT-like aminoacylation activity (Antonellis et al., 2003(Antonellis et al., , 2006Nangle et al., 2007;Niehues et al., 2015). Given the differential enzymatic activities of GlyRS variants, it is not surprising for scientists to seek out other diseaseassociated functions of these variants that are distinct from the canonical aminoacylation functions.

| Animal models of GlyRS-linked CMT
Among all CMT-associated aaRSs, GlyRS is well-established and the most studied in animal models. In Drosophila models, GlyRS variants significantly reduce the levels of newly synthesized proteins, and this translational defect is not attributed to altered tRNA aminoacylation because overexpression of Drosophila gars fails to rescue impaired protein translation (Niehues et al., 2015). Additionally, the enzymatic activity in the tissues of mice carrying heterozygous Gars Nmf249/+ and Gars C201R/+ variants (corresponding to GARS P234KY and GARS C157R in human) was not significantly decreased compared with that in WT animals (Achilli et al., 2009;Seburn et al., 2006).
Notably, overexpression of WT GlyRS could not improve the neuropathy phenotypes in either Gars Nmf249/+ or Gars C201R/+ mice (Motley et al., 2011), and the same conclusion regarding another GARS G240R variant was reached with Drosophila CMT models, which showed no phenotypic rescue upon overexpression of the WT proteins (Niehues et al., 2015). F I G U R E 1 Distribution and conservation of CMT-associated variant sites in human GlyRS. (a) Functional domains of GlyRS including a WHEP domain (in purple), a catalytic domain (in grey) and an anticodon domain (in yellow). (b) The crystal structure of human GlyRS (PDB entry 2PME). CMT variant sites in either schematic diagram (a) or crystal structure (b) are indicated with different colors based on enzymatic activity of each variant. CMT variants with WT-like enzymatic activity (fully active) are colored in green; variants with the activity ≥1/2 are labeled in blue; variants with the activity <1/2 are indicated in pink; variants with no activity (inactive) are colored in red; variants with activity undetermined are indicated in black. The priority of enzymatic activity displayed here is ranked based on aminoacylation assays in patient sample > animal models > in vitro using purified human enzyme > yeast orthologs. (c) Evolutionary conservation analysis of CMT-linked GlyRS across archaea, bacteria and eukaryotes. Sequence alignment of each GlyRS variant site is indicated by the color intensity, with blue representing variable and red representing conserved WT (Antonellis et al., 2003(Antonellis et al., , 2006Nangle et al., 2007;Niehues et al., 2015;Qin et al., 2014;Sivakumar et al., 2005) L74R P6 in F1  (Antonellis et al., 2003(Antonellis et al., , 2006He et al., 2011;Nangle et al., 2007;Sivakumar et al., 2005) D146N P2 in F1     Average age at onset.
c Average age at onset of available individual(s) except those deceased individuals, asymptomatic individuals, affected individuals with age at onset unknown or ambiguous, or alive individuals but refused to have a clinical evaluation.
g Difference of radius of gyration (R g , Å) for each variant relative to WT enzyme based on SAXS analysis. The value is calculated using the formula of (R

TA B L E 1 (Continued)
Recently, one group identified a de novo GARS variant (GARS ΔETAQ ) in a single patient with severe peripheral neuropathy, and this allele was further introduced into a mouse model (Morelli et al., 2019). This mouse model carrying a human disease allele displays primary features of peripheral neuropathy.
Strikingly, the allele-specific knockdown of GARS ΔETAQ using RNAi prevents the neuropathy phenotypes in mice, and the same efficacy is confirmed in Gars Nmf249/+ mice, both before and after onset.
These findings provide important proof-of-concept for virally delivered RNAi-based gene therapy for treating CMT2D; however, whether this approach can be applied to other CMT-causing single-base pair variants requires additional research. Of note, in this study, re-evaluation of in vitro kinetic properties and yeast models for the P234KY allele supported a loss-of-function effect (Morelli et al., 2019), contrary to previous reports (Table 1) Seburn et al., 2006;Stum et al., 2011). Such discrepancies in activity assays may reflect different sensitivities of experimental settings.
Given the well-characterized Gars Nmf249/+ in mice, one group further modeled the P234KY variant in Drosophila (gars P234KY ).
Ubiquitous expression of gars P234KY was shown to severely affect fly fitness, with no adults emerging, whereas pan-neuronal expression caused late pupal lethality, suggesting that toxicity may manifest or derive from tissues beyond the nervous system (Ermanoska et al., 2014). Indeed, muscular expression of gars P234KY in flies induces significant neuronal defects, and the same is true for the gars G240R variant, suggesting that the pathology has a noncell autonomous contribution (Grice et al., 2015). Interestingly, the neuronal toxicity of gars P234KY is dependent on the WHEP domain since its removal from gars P234KY abrogates neuromuscular and survival defects, revealing a clear dominant toxic gain-of-function mechanism for variants that may contribute to the pathology of CMT2D (Grice et al., 2015).
In contrast, mice carrying homozygous GARS variants, such as those of Gars Nmf249/Nmf249 and Gars C201R/C201R , are not viable or even undergo early death, and perinatal lethality can be rescued by overexpression of the WT proteins (Achilli et al., 2009;Seburn et al., 2006
Nonetheless, different CMT-causing variants have distinct effects on dimer formation. For example, GARS D500N , GARS G526R and GARS S581L strengthen the capacity of dimer formation, whereas GARS L129P and GARS G240R disrupt dimer formation . Because the dimeric form of GlyRS is essential for aminoacylation, both GARS L129P and GARS G240R show a loss of enzyme activity, while GARS D500N and GARS S581L demonstrate full aminoacylation activity compared to that of the WT enzyme. However, despite an enhanced dimer formation capability, GARS G526R has abolished aminoacylation activity . Therefore, considering this together with another fully active variant (GARS E71G ), almost half of these CMT-causing variants are active, further supporting the conclusion that CMT is not simply caused by a deficiency in aminoacylation. Furthermore, this finding also raises the possibility that CMT-causing variants may disrupt an unknown function of GlyRS, leading to a toxic gain-of-function that is associated with only GlyRS variants.
Considering these data, the same research group further ex- Whether the induced conformational changes confer the variants with the ability to interact with other proteins needs to be further addressed.
It is important to note that, based on conservation analysis (Figure 1c), some variant sites such as E71 and G526 in GARS, show less divergence in the evolutionary progression to humans, suggesting essential roles of these sites during evolution; however, based on the above findings, variations in these sites do not significantly affect enzymatic activity, further supporting non-canonical activities outside of aminoacylation imparted by these variants. Other essential functions of aaRSs beyond translation have been well documented (Guo & Schimmel, 2013;). in Gars Nmf249/+ mice, with specific targeting at peripheral nerves rather than the brain or spinal cord. More essentially, the defective phenotypes can be rescued by the HDAC6 inhibitor, and the same is true for Gars C201R/+ mice from another study (Benoy et al., 2018).

| Interaction partners of GARS variants
Although abnormal GlyRS-HDAC6 interactions have been identified in many human GARS CMT2D variants, the differential degrees of interplay among variants is closely associated with distinct clinical manifestations among CMT2D patients. For example, the GARS S581L and GARS G598A variants trigger the strongest HDAC6 interaction among all variants, of which the GARS G598A variant can also interact with Nrp1, concordant with unfavorable clinical features of infantile onset and extreme severity for patients harboring the GARS G598A variant (Mo et al., 2018). For the other human GARS CMT2D variants with relatively weak HDAC6 interactions, the possibility of alternative pathogenic mechanisms in CMT2D cannot be excluded.

| Mitochondrial role of GlyRS in CMT
It is worth noting that GlyRS is one of two aaRSs (the other is lysyl-tRNA synthetase) encoding both cytosolic and mitochondrial forms;

| A ARS VARIANTS
AlaRS is the third aaRS known to be involved in CMT. Unlike other CMT-associated aaRSs, AlaRS does not have an anticodon domain but rather has an editing domain and a C-terminal domain C-Ala ( Figure 3a). To date, nine AARS variants leading to dominant axonal CMT type 2N (CMT2N, OMIM# 613287) have been reported worldwide (Bansagi et al., 2015;Latour et al., 2010;Lin et al., 2011;McLaughlin et al., 2012;Motley et al., 2015;Weterman et al., 2018;Zhao et al., 2012) (Table 1)  variants, the AARS R329H variant was first discovered to segregate with CMT2N in two unrelated French families (Latour et al., 2010) and was then identified in a large Australian family (McLaughlin et al., 2012) and in a cohort of patients from one Irish and four British families (Bansagi et al., 2015), thus representing a recurrent variant worldwide. Yeast complementation assay results revealed impaired enzyme functions of the AARS R329H variant as well as the AARS N71Y , AARS G102R , AARS R326W and AARS S627L variants (McLaughlin et al., 2012;Motley et al., 2015;Weterman et al., 2018). In contrast, the cellular growth of yeast was not shown to be affected by the AARS E778A variant compared to that of the WT enzyme (McLaughlin et al., 2012), and the AARS R337K variant even improved yeast cell growth and showed a nearly 4-fold increase in tRNA charging activity (Weterman et al., 2018). Interestingly, although different enzyme functions are caused by the AARS R326W , AARS R337K and AARS S627L variants (Table 1), equivalent neural developmental toxicities were observed in the embryos of zebrafish after microinjections of human variant mRNAs, suggesting that the abnormal phenotypes in zebrafish are dominantnegative or toxic gains of function (Weterman et al., 2018).
Other variants in AARS, including AARS D893N from a Chinese family (Zhao et al., 2012) and AARS E688G from an Irish family (Bansagi et al., 2015), were identified; however, their impacts on enzymatic activity and phenotypes in animal models need to be verified in the future.
As mentioned above, AlaRS includes an editing domain wherein two human CMT-linked AlaRS variants (AARS S627L and AARS E688G ) are located (Figure 3a and b). In mice, a missense variant at A734E in the editing domain of murine AlaRS (Aars A734E ) can cause cell-lethal accumulation of misfolded protein in neurons, leading to severe neurodegenerative phenotypes . Although the A734E variant in Aars is recessive, this finding, to some extent, pinpoints the fundamental roles of the editing activity of AlaRS for maintaining the accurate processing of genetic information and provides insight into novel mechanisms underlying human neurodegenerative diseases, such as AlaRS-linked CMT.
Interestingly, the C-Ala domain was once termed the dimerization domain because it provides contacts for dimerization in archaeal AlaRS (Naganuma et al., 2009). Over evolutionary time, the C-Ala domain has been completely dispensable for aminoacylation but has developed distinct roles in higher organisms (Sun et al., 2016). Such a functional evolution of the C-Ala domain allows AlaRS to be the single exception among the 19 other human aaRSs with no new motif or domain additions . Perhaps for this reason, the AARS E778A variant in the C-Ala domain shows WT-like catalytic activity (McLaughlin et al., 2012), implying that nonenzymatic gain-of-function mechanisms underlie the pathogenesis of CMT2N.

| HARS VARIANTS
HisRS has a domain structure identical to that of GlyRS, as it consists of a WHEP domain, a catalytic domain and an anticodon binding domain (Figure 4a). HARS variants have been successively identified in dominant axonal CMT type 2W (CMT2W, OMIM# 616625) (Laura et al., 2019), with all variants being located in the catalytic domain ( Figure 4a and b). Of eight variants, seven result in a loss of function as determined by yeast complementation assays, and neurotoxicity has been successfully recapitulated in transgenic C. elegans models of HARS R137Q and HARS D364Y variants (Abbott et al., 2018;Safka Brozkova et al., 2015;Vester et al., 2013). However, the discrepancies between yeast model phenotype and aminoacylation activities in vitro (e.g., the HARS D364Y variant is lethal in yeast cells but has normal charging activity in vitro) (Table 1), as well as the causal link between these variants and CMT2W, remain unclear.
Based on these considerations, a recent study further investigated charged tRNAs in actual patients with CMT using assays enabling the detection of aminoacylation within the human cell    and HARS D364Y variants in CMT are also rationalized by conservation analysis which showed highly conserved P134 and D364 sites with full enzymatic activity (Figure 4c).
It is worth noting that both GlyRS and HisRS contain a WHEP domain, which is indispensable for the neuronal toxicity caused by the gars P234KY variant in Drosophila (Grice et al., 2015). WHEPmediated suppression has also been shown to be a gain of function in the GARS G526R variant (He et al., 2011). In the case of HisRS, the conformational changes induced by HisRS variants strengthen the interactions between the WHEP domain and the C-terminal part of the catalytic domain, which may help to open the dimerization interface . Interestingly, removal of the WHEP domain mostly had no effect on the aminoacylation activities of the WHEP-containing aaRSs , suggesting a potential relevance of the WHEP domain for CMT.  (He et al., 2015). In this regard, axonal neuropathy is very likely to be closely associated with the angiogenesis pathway, but this nonenzymatic function needs to be further investigated.

| MARS variants
Unlike other CMT-linked aaRSs, methionyl-tRNA synthetase (MetRS) functions as a monomer and associates with the multi-synthetase complex (MSC), which is anchored by the N-terminal GST domain.
MetRS also has a conserved class I catalytic domain and an anticodon domain, followed by a C-terminal-appended WHEP domain with an unclear function. Four MARS variants have thus far been linked to CMT type 2U (CMT2U, # OMIM 616280) (Table 1). Three variants are located in the anticodon binding domain, and one is in the catalytic domain. The monoallelic variant of R618C in the MARS (MARS R618C ) was first described in a family with two affected patients who presented with late-onset CMT2U and one unaffected member (Gonzalez et al., 2013). The MARS R618C variant is nonfunctional in yeast models, suggesting a loss-of-function feature; however, the potential mechanisms and contributions of this variant in CMT are less well characterized (Gonzalez et al., 2013). Subsequently, the MARS P800T variant was identified in a Korean family with late-onset CMT2U (Hyun et al., 2014), and then in another Japanese family with late-onset CMT2U (Hirano et al., 2016) and a Korean family with CMT2U but with a relatively early onset (Nam et al., 2016), suggesting that this recurrent variant causes variability and diversity of the CMT phenotype. Recently, two novel, likely disease-associ-

| NARS variants
In addition to CMT, aaRSs variants have been frequently implicated in other neuropathies. For example, biallelic NARS variants were identified in 7 affected patients with recessive microcephaly from three unrelated families (Wang et al., 2020). Another study described de novo dominant heterozygous and biallelic recessive variants in the NARS in 32 individuals from 21 families, presenting with multiple neurodevelopmental defects (Manole et al., 2020).
Interestingly, functional data indicated that genotypes with dominant heterozygous NARS variants produce a toxic gain-of-function, whereas the homozygous recessive variants probably induce a partial loss of function (Manole et al., 2020;Wang et al., 2020). Although neuropathies other than CMT are outside the scope of this review, these studies do shed light on the complex pathogenic mechanisms of aaRSs in neuropathies.

| SUMMARY AND PROS PEC TS
Despite the identification of an increasing number of causative genes, CMT remains incurable. A more accurate classification and deeper understanding of CMT remain great challenges to scientists.   Seburn et al., 2006), suggesting that some variants result in undefined loss of function. Nevertheless, the specific mechanism by which aminoacylation deficiency causes peripheral neuropathy remains unknown.
Many lines of evidence have confirmed the toxic functions gained from aaRS variants, and this discovery may serve as a shared disease-causing mechanism for aaRS-associated CMT. For example, different CMT-linked variants of GlyRS, HisRS and TyrRS lead to shared neomorphic structural opening which allows aberrant interactions with membrane receptors or intracellular proteins, thereby interfering with certain signaling pathways and trafficking in motor and sensory neurons (Blocquel et al., 2017He et al., 2011He et al., , 2015. This shared property may facilitate the identification of new therapies by targeting these opened sites in all aaRS-linked CMT subtypes. In addition, both CMT2D and DI-CMTC models in Drosophila share common genetic modifiers with nuclear localization (Ermanoska et al., 2014). This finding implies that the nuclear involvement of CMT-linked aaRSs, such as TyrRS (Bervoets et al., 2019), may very likely be another shared pathogenic mechanism, but this topic needs further exploration. Last, except for TyrRS and AlaRS, three CMT-linked aaRSs (GlyRS, HisRS and TrpRS) and one putative CMTassociated MetRS contain a specific appended WHEP domain, which mostly does not affect the enzymatic activities of their aaRSs  but does show a close association with aaRSlinked CMT Grice et al., 2015;He et al., 2011), implying a potential shared pathogenic mechanism. Further studies on the specific role of the WHEP domain in CMT are of great interest. Nonetheless, we cannot rule out the possibility that both functional losses and functional gains are simultaneously involved in the pathogenic mechanisms in certain CMT forms, although the underlying factors remain unknown.
Furthermore, multiple aaRS variants have been identified in CMT, but only a few have been tested in animal models or patient cells. To date, the CMT-causing GlyRS is the only one that has been successfully recapitulated in mouse models, while the functions of the other aaRS variants have been determined in only yeast strains, which are apparently insufficient to illustrate the true regulatory functions of aaRSs under physiological conditions. The discrepant aminoacylation levels of the HARS P134H variant in yeast models and CMT patients suggests that the pathogenic mechanisms caused by aaRS variants in CMT are highly context-dependent. Furthermore, simple experimental models of aaRS variants established in flies, fish and worms also require re-evaluation in mammalian animal models or patient cells. As such, detailed analysis of a higher model system will be critical for addressing this issue, and it will help to not only distinguish the contributions of each aaRS variant to CMT, but also to determine the pathogenic commonalities among different aaRS-associated CMT subtypes to further advance therapeutic interventions.
Last but not least, neither ideal biomarkers nor therapeutic targets are currently available for slowly progressive CMT, although we do have some promising candidates to target in the diagnosis and treatment of CMT, such as Nrp-1, TRIM28 and nuclear TyrRS.
Deeply understanding why peripheral nerves are predominantly affected in aaRS-linked CMT and how they work within the human cell environment will hopefully lay the foundation for the precise stratification and targeted treatment of CMT in the future.

CO N FLI C T S O F I NTE R E S T
The authors report no conflicts of interest.

AUTH O R S' CO NTR I B UTI O N S
H.Z. and L.S. wrote the paper and made the figures. Z-W.Z. and L.S. designed the review.