Ubiquitously expressed proteins and restricted phenotypes: exploring cell-specific sensitivities to impaired tRNA charging

Abstract

Aminoacyl-tRNA synthetases (ARS) are ubiquitously expressed, essential enzymes that charge tRNA with cognate amino acids. Variants in genes encoding ARS enzymes lead to myriad human inherited diseases. First, missense, nonsense, and frameshift alleles cause recessive multi-system disorders that differentially affect tissues depending on which ARS is mutated. Second, missense alleles cause dominant peripheral neuropathy. A preponderance of evidence has shown that both phenotypic classes are associated with loss-of-function alleles, suggesting that tRNA charging plays a central role in disease pathogenesis. However, it is currently unclear how perturbation in the function of these ubiquitously expressed enzymes leads to tissue-specific or tissue-predominant phenotypes. Here, we review our current understanding of ARS-associated disease phenotypes and explore potential explanations for the observed tissue specificity.

Ubiquitously expressed proteins and tissue-specific effects

Ubiquitously expressed genes typically encode proteins that are required for all cell types [1]. Interestingly, variants in ubiquitously expressed, essential genes sometimes result in tissue-specific or tissue-predominant phenotypes in affected patients. For example, variants in 20 ubiquitously expressed ribosomal protein genes have been implicated in the blood disorder Diamond-Blackfan anemia (MIM 105650), which is a defect of erythroid progenitors [2, 3]. Similarly, variants in the ubiquitously expressed tRNA nucleotidyl transferase 1 locus (TRNT1 [MIM 612907]) [4] have been implicated in phenotypes affecting erythrocytes including sideroblastic anemia [5] and erythrocytic microcytosis [6]. These observations suggest that some cells are more sensitive to defects in housekeeping processes, or that some essential genes have secondary functions that are critical for a subset of tissues. Recent findings have shown that variants in aminoacyl-tRNA synthetases—a family of enzymes critical for protein translation in all cells—cause myriad clinical phenotypes that vary depending on the affected enzyme. Here, we explore tissue-specific phenotypes associated with aminoacyl-tRNA synthetase gene variants and provide plausible mechanisms to explain these observations.

Aminoacyl-tRNA synthetases and human inherited disease

Aminoacyl-tRNA synthetases (ARSs) are ubiquitously expressed, essential enzymes that are responsible for charging tRNA with cognate amino acids, a crucial step in translating genetic information into proteins [7]. ARSs charge tRNA through a two-step reaction that involves: (i) activating the amino acid with ATP to form an aminoacyl adenylate intermediate; and (ii) transferring the aminoacyl group to the tRNA and releasing the charged tRNA (Figure 1) [7]. Each ARS enzyme contains peptide sequences that are important for activating the amino acid (e.g., a catalytic domain) as well as sequences important for recognizing the appropriate tRNA molecules (e.g., an anticodon binding domain) [8]. ARS activity is specific to amino acid and tRNA pairings, and there are multiple mechanisms that contribute to this specificity including recognition of amino acids by the ARS active site and molecular signals in tRNAs that trigger aminoacylation [9, 10]. In addition to tRNA charging, secondary functions have been described for ARS enzymes [11, 12]. For example, ARSs participate in inflammation, angiogenesis, apoptosis, and tumorigenesis [11, 12]. This expansion of functions has been linked to the addition of appended domains [13] and to protein isoforms arising from alternative splicing [14, 15]. In total, there are 37 ARS genes encoded in the human nuclear genome: 17 encode cytoplasmic enzymes; 17 encode mitochondrial enzymes; and three encode enzymes that function in both compartments [7]. ARSs are abbreviated using the single-letter code for the amino acid followed by ‘ARS1’ for cytoplasmic enzymes or ‘ARS2’ for mitochondrial enzymes [7]. For example, cytoplasmic cysteinyl-tRNA synthetase (MIM 123859) is abbreviated ‘CARS1’, and mitochondrial cysteinyl-tRNA synthetase (MIM 612800) is abbreviated ‘CARS2’.

Figure 1.

The two-step aminoacylation reaction. (A) First, the aminoacyl-tRNA synthetase (ARS) binds the amino acid (AA) and an ATP molecule to form the aminoacyl adenylate intermediate; this process results in the release of a pyrophosphate molecule. (B) Second, a cognate tRNA molecule binds the ARS and the amino acid is transferred to the tRNA; an AMP molecule and the charged tRNA are then released. The image represents a dimeric enzyme where each subunit is capable of charging tRNA.

Over the past 15 years, variants in ARS loci have been associated with a variety of human inherited disease phenotypes (Table S1). In 2003, variants in glycyl-tRNA synthetase 1 (GARS1 [MIM 600287]) were reported in patients with dominant peripheral neuropathies (also referred to as Charcot-Marie-Tooth [CMT] disease) [16]. Since then, variants in four additional ARS loci have been implicated in CMT disease [17-20]. The mechanism by which ARS variants affect the peripheral nervous system has remained elusive. Additionally, variants in 34 of the 37 ARS loci have been implicated in early-onset, multi-system recessive disease phenotypes [21]. In these cases, most patients present with overlapping clinical phenotypes; however, some tissues appear to be more sensitive to the reduced function of a particular ARS and the mechanisms behind this tissue specificity are not defined. Here, we review our current understanding of ARS-related inherited diseases and propose a path forward toward defining the mechanisms that lead to tissue-specific or tissue-predominant phenotypes.

Dominant peripheral neuropathies: tRNA charging deficits and axon function

Inherited or de novo variants in five ARS genes (GARS1, YARS1 [MIM 603623], AARS1 [MIM 601065], HARS1 [MIM 142810], and WARS1 [MIM 191050]) have been implicated in dominant Charcot-Marie-Tooth disease (CMT), an inherited neurodegenerative peripheral neuropathy that impairs motor and sensory function in the distal extremities [16-20]; please note that while variants in MARS1 (MIM 156560) have been reported in patients with dominant CMT [22-26], a genetic argument for pathogenicity has not been established [22-24]. Neuropathy-associated ARS variants are uniformly missense or in-frame deletion variants [17]. The majority of these variants show loss-of-function effects in functional assays including in vitro aminoacylation assays and yeast complementation assays [21, 27]. Though data indicate loss-of-function effects for dominant-neuropathy-associated alleles, haploinsufficiency is unlikely to be the disease mechanism based on: (i) studies showing that heterozygous null Gars1 mice do not have a neuropathic phenotype while mice heterozygous for pathogenic missense variants do [28, 29]; and (ii) the frequency of null alleles in the heterozygous state in gnomAD for each of the five ARS genes implicated in dominant neuropathy [30]. Indeed, the probability of loss of function intolerance (pLI) has been calculated for these five genes based on the number of loss-of-function variants identified in gnomAD (pLI is 0 for YARS1, AARS1, and HARS1; 0.3 for WARS1; and 0.31 for GARS1), which predicts that these genes are tolerant to the loss of one allele (the closer the pLI is to one, the more intolerant the gene is to the loss of one allele), further indicating that haploinsufficiency is unlikely [30].

Two non-mutually-exclusive hypotheses have been proposed to explain the mechanism of ARS variants in dominant neuropathy (Figure 2). First, since all five implicated enzymes function as homodimers and since the variants show loss-of-function effects, a dominant-negative mechanism may explain the dominant phenotype (Figure 2A) [21]. Non-functional mutant subunits may dimerize with wild-type subunits in a heterozygous patient cell and drastically reduce tRNA charging. An impairment in ARS function—and therefore protein translation—may be particularly detrimental to long peripheral nerve axons, potentially explaining the tissue-specificity [21]. Further work investigating if dimerization is necessary for pathogenicity and if variants in monomeric enzymes can cause dominant neuropathy will help assess the role of dominant-negative effects in disease pathogenesis.

Figure 2.

Potential mechanisms of ARS-associated dominant peripheral neuropathy. (A) Non-functional mutant ARS subunits may dimerize with wild-type subunits in a heterozygous patient cell and drastically reduce tRNA charging—and subsequently protein translation—via a dominant-negative effect. (B) Mutant ARS enzymes may aberrantly bind NRP1, TRK, or HDAC6 and perturb neuronal signaling pathways. A key is provided to define each component; see text for details.

A second hypothesis states that gain-of-function effects contribute to disease pathogenesis (Figure 2B) [21, 31]. Variants in GARS1 [32] and YARS1 [33] induce conformational openings that permit aberrant interactions with other proteins [31]. Mutant GARS1 has been reported to aberrantly interact with neuropilin 1 (NRP1 [MIM 602069]) [34], tropomyosin receptor kinase (TRK) receptors [35], and histone deacetylase 6 (HDAC6 [MIM 300272]) [36, 37]. Similarly, mutant YARS1 has been reported to aberrantly interact with tripartite motif-containing protein 28 (TRIM28 [MIM 602742]) [33]. These interactions have been proposed to explain the sensitivity of the peripheral nervous system to GARS1 and YARS1 variants via neuron-specific pathways [31]; however, it should be noted that interactions between GARS1 and NRP1 do not explain early-onset, severe, GARS-associated neuropathy [38]. Regardless of the molecular mechanism, it is clear that the pathogenic alleles are dominantly toxic. In support of this notion, allele-specific RNAi rescues aspects of dominant neuropathy in a mouse model of GARS1 variants [38], which is consistent with dominant-negative and gain-of-function mechanisms.

Variants in tRNA synthetase genes cause a spectrum of recessive disease phenotypes

To date, genetic variants in 34 of the 37 ARS loci have been implicated in human recessive disease phenotypes [21]; these 34 loci include all 17 genes encoding an ARS enzyme that functions only in the mitochondria [21]. Unsurprisingly, disease-associated variants affecting mitochondrial ARS enzymes typically cause clinical manifestations in highly metabolic tissues. These include, but are not limited to, the central nervous system (CNS) (CARS2 [39], FARS2 [MIM 611592] [40], NARS2 [MIM 612803] [41, 42], PARS2 [MIM 612036] [41, 42], RARS2 [MIM 611524] [43], TARS2 [MIM 612805] [44], DARS2 [MIM 610956] [45], EARS2 [MIM 612799] [46-49], MARS2 [MIM 609728] [50, 51], and WARS2 [MIM 604733] [52]), muscle (YARS2 [MIM 610957] [53]), ovaries (HARS2 [MIM 600783] [54], LARS2 [MIM 604544] [55], and AARS2 [MIM 612035] [56]) and the skeletal system (IARS2 [MIM 612801] [57]) [21, 58]. Diseases associated with mitochondrial ARSs have been classified based on whether variants cause clinical manifestations in the CNS and/or other tissues, highlighting the spectrum of tissues involved [58]. It is currently unclear why variants in mitochondrial ARSs cause phenotypes affecting different tissues depending on which enzyme is affected. Two models for explaining this phenomenon have been presented [58]. First, since mitochondrial protein translation is important for oxidative phosphorylation, the tissue-specific phenotypes may arise due to temporal and spatial differences in energy requirements [58]. Second, mitochondrial ARSs may have secondary functions that have yet to be discovered and that may be affected by the pathogenic variants [58]. In addition, amino-acid concentrations and mitochondrial tRNA expression may be variable in different tissues [59], which would convey differential susceptibility to impaired mitochondrial ARS function. Further work investigating tissue-specific energy requirements, secondary functions of mitochondrial ARS enzymes, and the basic biology of tRNA expression and amino-acid availability in mitochondria is needed to address these questions.

Variants in 17 of the 20 ARS loci that encode cytoplasmic enzymes have been implicated in recessive diseases that often affect a wide array of tissues [21]. Pathogenic alleles in these 17 ARS genes largely cause overlapping recessive clinical presentations, which often include neurodevelopmental phenotypes. Interestingly, some tissues appear to be more sensitive to defects in a particular ARS gene. Extensive genetic and functional data implicate a loss-of-function molecular pathology in ARS-associated recessive disease [21]. Given that all of the ARSs share the common function of tRNA charging, a logical notion would be that disease-associated ARS variants cause defects in protein translation, which then contribute to disease pathogenesis (Figure 3A, Key Figure). However, while a preponderance of evidence points to reduced ARS function as the culprit in ARS-related recessive phenotypes, the role of secondary functions cannot be ruled out (Figure 3B).

Figure 3.

Potential mechanisms of ARS-associated recessive disease. (A) Wild-type CARS1 (left) charges tRNA^Cys with cysteine, and ribosomes are then able to translate transcripts with many cysteine codons. In contrast, mutant CARS1 (right) has reduced charging activity, which could result in uncharged tRNA^Cys and the stalling of ribosomes along transcripts with many cysteine codons (indicated by small curved lines around ribosome). The impaired charging activity may be exacerbated by reduced levels of tRNA^Cys (grayed out tRNA) and/or cysteine (grayed out circles) in specific tissues. (B) Wild-type LARS1 (left) is an activator of mTORC1, which inhibits autophagy. Activation of mTORC1 may be impaired by mutant LARS1 (right), which could lead to abnormally increased autophagy. A key is provided to define each component; see text for details.

While loss-of-function ARS alleles have been convincingly implicated in a range of recessive phenotypes, a critical question has not been addressed: Why do some tissues appear to be more sensitive to impairments of particular ARSs? We speculate that differing requirements across tissues to translate proteins enriched for specific amino acids may contribute to the tissue-specific phenotypes. That is, some tissues may have high demands for proteins rich in a certain amino acid and therefore may be more sensitive to defects in charging tRNA with that amino acid. In addition, it is known that different tissues have differing availabilities of specific tRNAs [59, 60] and amino acids [61, 62], which could also influence whether a tissue is more or less sensitive to defects in specific ARSs. Here, we explore potential mechanisms to explain tissue-specific features of ARS-related disease using examples of ARS loci that cause recessive phenotypes (Figure 4).

Figure 4.

Variants in ARS genes have been associated with tissue-specific or tissue-predominant recessive phenotypes. Variants in CARS1 and TARS1 have been implicated in brittle hair (blue), and CARS1 variants have also been implicated in brittle nails (blue). Variants in KARS1 have been associated with sensorineural hearing impairment (red). Variants in MARS1 have been implicated in interstitial lung disease and pulmonary alveolar proteinosis (yellow). Variants in LARS1 have been implicated in infantile hepatopathy (green).

Brittle hair and nails: cysteine, threonine, and keratin biology

Hair is a keratinized tissue that contains two major structural components: (i) hair keratins that form keratin intermediate filaments; and (ii) keratin-associated proteins that are located in the matrix around—and form disulfide bonds with—the keratin intermediate filaments [63]. The nail plate is another keratinized structure that also has a high content of keratin proteins and keratin-associated proteins [64, 65]. Variants in genes encoding hair and nail keratins result in diseases affecting the structure of these tissues including monilethrix (MIM 158000) and ectodermal dysplasia of hair and nail type [66, 67]. Variants in two ARS loci, cysteinyl-tRNA synthetase 1 (CARS1) and threonyl-tRNA synthetase 1 (TARS1; MIM 187790), cause brittle hair and nails, suggesting that keratinocytes may be particularly sensitive to impaired tRNA charging with cysteine and threonine.

CARS1 encodes the enzyme that charges tRNA^Cys with cysteine in the cytoplasm [7] and bi-allelic variants in CARS1 cause a multi-system, recessive disease [68]. CARS1 variants were identified in four patients from three families, and each of the variants showed loss-of-function effects on CARS1 protein expression, tRNA charging, or cell viability in a yeast model [68]. The patients presented with features that overlap with other ARS-related recessive phenotypes including developmental delay, microcephaly, and liver dysfunction [68]. Interestingly, all patients also presented with brittle hair and nails [68], features that had not been reported in ARS-associated diseases until recently when patients with TARS1 variants were reported with brittle hair [69]. In the latter study, two patients were diagnosed with trichothiodystrophy (TTD), a rare, recessive, neuroectodermal disease with characteristic brittle hair [69]. Each patient carried bi-allelic TARS1 variants that cause loss-of-function effects on TARS1 protein levels and tRNA charging [69]. One of the patients additionally had ichthyosis, follicular keratosis, delayed physical development, recurrent respiratory tract infections, and acromandibular dysplasia [69]. The second patient was born encased in a tight membrane and had ichthyosis [69]. A nail phenotype was not reported in these patients [69].

How CARS1 variants lead to brittle hair and nails, and how variants in TARS1 lead to brittle hair is not understood. For CARS1 variants, this phenotype may reflect the high cysteine requirements of hair and nails [68], since hair and nail keratins and keratin-associated proteins contain many cysteine residues that participate in disulfide bonding. In fact, we performed a computational analysis of the human proteome to identify proteins with cysteine content above the proteome average (3%) and found that 25 of the top 30 proteins with the highest percentages of cysteine (35-41%) are keratin-associated proteins (unpublished data, Table 1). Additionally, gene ontology analysis of genes encoding proteins with higher than average cysteine content showed an enrichment for gene products associated with ‘keratinization’ (GO:0031424; p-value = 4.35e-28) (Table S2). Since disease-associated CARS1 variants cause a loss-of-function effect, tRNA^Cys charging is likely impaired, which may lead to defects in cysteine-rich protein translation, including keratins and keratin-associated proteins; such an effect may explain the brittle hair and nail phenotype. For TARS1, it has been suggested that pathogenic variants cause brittle hair due to reduced protein translation [69]; however, this does not explain why this phenotype is only observed in patients with TARS1 and CARS1 variants. Though keratin-associated proteins are not the most threonine-rich proteins in the proteome, many cysteine-rich, keratin-associated proteins also have threonine content higher than the proteome average (5%). For example, keratin-associated protein 9-2 (KRTAP9-2), which is 33% cysteine, is also 16% threonine. Gene ontology analysis of genes encoding proteins with higher than average threonine content showed an enrichment for gene products associated with adhesion including ‘cell-cell adhesion’ (GO:0098609; p-value = 1.83e-10) (Table S2), which is an important process for keratinocytes [70]. Further work is needed to determine if cysteine- and threonine-rich proteins are affected by CARS1 and TARS1 variants. Our analyses revealed another group of proteins that have high cysteine content: metallothioneins (MTs), which are metal-binding proteins that function in detoxification of heavy metals such as cadmium and copper [71]. If MTs are affected by disease-associated CARS1 variants, patients may be more susceptible to heavy metal toxicity. Experiments in relevant models in the environment of depleted CARS1 function are needed to evaluate the effect on metallothionein protein levels and to test for differential sensitivity to heavy metal exposure.

Table 1.

Top 30 human proteins with the highest percentages of each indicated amino acid.

Cysteine		Threonine		Lysine		Leucine
Protein	% Cys	Protein	% Thr	Protein	% Lys	Protein	% Leu
SCYGR5	41%	IGHD1-1	60%	TMA7	31%	PPY2P	38%
SCYGR1	40%	MUC2	35%	H1F0	29%	MYMX	30%
SCYGR3	39%	MUC22	34%	HIST1H1B	29%	SFTA2	26%
SCYGR7	39%	MUC3A	29%	HIST1H1E	28%	MFSD3	25%
SCYGR6	38%	TRBJ2-3	25%	RPL41	28%	OCLM	25%
SCYGR4	37%	TRBD1	25%	HIST1H1C	28%	MT-ND3	24%
KRTAP4-6	37%	MUC17	24%	HIST1H1D	28%	TTTY13	24%
SCYGR8	37%	MUC12	22%	H1-1	27%	SCGB3A2	24%
KRTAP4-12	37%	MUC3B	21%	SREK1IP1	26%	MT-ND4L	23%
KRTAP4-4	37%	MUC5B	21%	CYLC2	26%	TMEM216	23%
SCYGR2	37%	MUC5AC	20%	FAM133B	25%	GLYCAM1	23%
KRTAP4-5	36%	PRG4	20%	FAM133A	25%	TSPO2	23%
KRTAP4-3	36%	TRBJ2-7	20%	HMGN2	24%	GP1BB	23%
KRTAP4-7	36%	TRBJ1-2	20%	HMGN4	23%	TMEM203	23%
KRTAP17-1	36%	TRBJ1-1	20%	LLPH	23%	SMIM40	23%
KRTAP5-7	36%	MUC19	20%	RPL38	23%	SLC39A5	23%
KRTAP9-1	36%	MUC21	19%	CYLC1	23%	TMEM82	23%
KRTAP5-9	36%	TRGJ1	19%	FAU	22%	BPIFA1	23%
KRTAP4-11	35%	TRDJ1	19%	TRDN	22%	IL11	23%
KRTAP5-8	35%	HCG22	19%	RPS25	22%	SLC35A4	23%
KRTAP5-11	35%	HAVCR1	18%	RPL39P5	22%	NRM	23%
KRTAP4-9	35%	KLF18	18%	FAM32A	21%	SFT2D3	22%
KRTAP4-8	35%	MUCL1	18%	C19orf53	21%	IGFALS	22%
KRTAP5-6	35%	MUC16	18%	RPL35	21%	TMEM35B	22%
KRTAP4-2	35%	TRBJ1-6	18%	HMGN1	21%	CYB561D2	22%
MT1B	34%	PLAC4	17%	RPL36A	21%	BIK	22%
KRTAP4-1	34%	MUC6	17%	RPL13AP3	21%	IL27	22%
KRTAP5-3	33%	TMEM123	17%	TMSB4X	20%	LRRC32	22%
MT1F	33%	MTRNR2L9	17%	SERF2	20%	TMEM140	22%
MT1E	33%	MTRNR2L6	17%	RPL29	20%	MYMX	30%

The human proteome was obtained from UniProt (Proteome ID UP000005640; accessed November 3, 2019). Proteins from the Swiss-Prot UniProt database are shown and uncharacterized proteins were removed. Biostrings in R was used to determine the number of each amino acid in each protein. For each protein, percent for a given amino acid was calculated by dividing the number of that specific amino acid by the total number of amino acids. Here we show the top 30 proteins for each of the listed amino acids. See also online supplemental information Table S2.

Sensorineural hearing loss: lysine and the inner ear

The inner ear has a complex cellular organization and, therefore, variants in a large panel of genes can cause hearing loss [72, 73]. Lysyl-tRNA synthetase 1 (KARS1; MIM 601421) encodes the enzyme that charges tRNA^Lys with lysine in both the cytoplasm and the mitochondria [74]. Interestingly, bi-allelic KARS1 variants have been associated with recessive, non-syndromic hearing impairment [73], suggesting that inner ear cells may be particularly sensitive to impairments in tRNA charging with lysine. In addition to non-syndromic hearing impairment, KARS1 variants have been implicated in recessive, syndromic hearing loss. In these latter patients, hearing loss is comorbid with other symptoms commonly associated with ARS variants including developmental delay, neurocognitive decline, seizures, and leukoencephalopathy [75-77].

It is not currently understood how the inner ear is particularly sensitive to mutations in KARS1. It has been suggested that aminoacylation impairments in the cochlea might affect many cellular processes and lead to hearing impairment [73]. A general sensitivity to impaired translation may explain why variants in other ARS loci (LARS2 [55], HARS2 [54], HARS1 [78], AARS1 [79], RARS2 [80]) have also been associated with broader syndromes that include hearing impairment, but it does not explain the non-syndromic hearing loss associated with KARS1 variants. We wondered if inner ear cells have high lysine requirements that may explain this tissue-specific phenotype. We analyzed the human proteome to identify proteins with percentages of lysine higher than the proteome average (6%) and then considered if those proteins had roles in the inner ear (unpublished data, Table 1). One protein of particular interest with high lysine content is growth arrest-specific 8 (GAS8 [MIM 605178]), which contains 12% lysine and is a component of the dynein regulatory complex that is important for ciliary motility and for ear development [81]. Similarly, the transmembrane inner ear expressed gene (TMIE [MIM 607237]) has been associated with autosomal recessive non-syndromic hearing loss [82] and encodes a protein with 15% lysine. Gene ontology analysis of genes encoding proteins with higher than average lysine content showed the greatest enrichment for gene products associated with protein targeting including ‘SRP-dependent cotranslational protein targeting to membrane’ (GO:0006614; p-value = 3.83e-8) and ‘protein targeting to ER’ (GO:0045047; p-value = 1.61e-8) (Table S2); however, it is unclear how perturbations in these processes would specifically affect the process of hearing. Future work is needed to investigate the roles of lysine-rich proteins in the inner ear and if the translation of lysine-rich proteins are specifically affected by KARS1 variants.

Infantile hepatopathy: leucine and liver function

Infantile liver disease is a life-threatening condition that develops in early infancy and can be difficult to diagnose [83]. Patients present with failure to thrive, jaundice, and distended abdomen, and may have elevated liver transaminases and hyperbilirubinemia [83]. Leucyl-tRNA synthetase 1 (LARS1) encodes the enzyme that charges tRNA^Leu with leucine in the cytoplasm [7]. Bi-allelic variants in LARS1 cause infantile liver failure syndrome type 1 (ILFS1; MIM 615438), a multi-system disorder involving infantile liver failure, developmental delay, anemia, seizures, and recurrent liver dysfunction triggered by febrile illness [83-86]. Notably, bi-allelic variants in many ARS genes result in phenotypes that affect the liver [21, 87], which may indicate that this organ is generally sensitive to impaired protein translation. However, LARS1 variants seem to predominantly affect the liver and cause earlier onset and a more severe liver phenotype; perhaps this liver-predominant effect may be due to impaired translation of proteins with high leucine content. It has been proposed that proteins with high levels of leucine were most likely to be affected in patients with LARS1 variants [83]. Previous analysis of the human proteome determined that immune-related proteins contain very high percentages of leucine, which correlates with the patient phenotype of hepatic dysfunction triggered with illness [83]. Additionally, leucine-rich proteins are involved in the phospholipid biosynthetic process and, in support of this observation, large fat deposits were seen in affected livers [83]. Our gene ontology analysis of genes encoding proteins with higher than average leucine content showed an enrichment for similar terms including ‘lipid biosynthetic process’ (GO:0008610; p-value = 1.79e-4) (Table S2). Furthermore, when analyzing the human proteome for leucine-rich proteins, we found that the gene associated with infantile liver failure syndrome 2 (MIM 616483) [88] encodes a protein with a higher percentage of leucine than the proteome average of 10%; NBAS (MIM 608025) is 13% leucine.

As mentioned above, loss of a secondary function could explain the tissue-specific effects of ARS variants and there are data suggestive of this for LARS1-associated disease. It has been reported that LARS1 is a leucine sensor and an activator of mTORC1 (an autophagy inhibitor), and that LARS1 deficiency results in reduced activation of mTORC1 and increased autophagy [89]. Reduced hepatic mTORC1 activity has been shown to cause liver cell damage; therefore, variants in LARS1 may contribute to disease through reduced mTORC1 activity and abnormal autophagy [84]. Aberrant mTORC1 signaling may also contribute to the seizure phenotype observed in these patients since mTOR signaling has been associated with neurological disorders including epilepsy [90]. Future research is needed to determine if leucine-rich protein expression is specifically affected by LARS1 variants, and if the disease-associated alleles affect mTORC1 activity.

Future directions toward defining the mechanisms of tissue-specific ARS phenotypes

Variants in ARSs have been implicated in a variety of human diseases with dominant and recessive inheritance patterns, and many questions on the mechanisms of disease pathogenesis remain unanswered (see Outstanding Questions). Variants in five ARSs have been implicated in dominant peripheral neuropathy, and both dominant-negative and gain-of-function mechanisms have been proposed to explain the peripheral nerve phenotype (Figure 2). Further work is needed to address critical questions about ARS-associated neuropathy including: Is there is a common, unifying feature among the five implicated enzymes that explains the dominant neuropathy? and Can mutations in any ARS enzyme cause dominant neuropathy or are there unique features about the five enzymes implicated to date?

OUTSTANDING QUESTIONS.

• What are the downstream consequences of impaired ARS function on protein translation? Does this result in a global defect in translation or does this cause a more dramatic impairment in the translation of a specific subset of proteins?

• Do disease-associated ARS alleles impact non-canonical functions and does this process play a role in human disease pathogenesis?

• For recessive diseases cause by ARS variants, are some tissues more affected than others due to a requirement for proteins enriched for the associated amino acid? Similarly, are some tissues more affected due to lower levels of the associated amino acid or tRNA?

• For ARS variants that cause dominant axonal neuropathy, does impaired enzyme function (possibly exacerbated by a dominant-negative effect) lead to the axonal dysfunction? Alternatively, is a toxic gain-of-function effect common to all neuropathy-associated ARS alleles? Finally, can variants any ARS enzyme cause dominant neuropathy, or is there something unique about the five that have been implicated to date?

As outlined above, variants in cytoplasmic ARSs cause recessive phenotypes often affecting a wide array of tissues; however, these variants have also been linked with tissue-specific or tissue-predominant effects and the mechanisms of the tissue specificity is not defined. Given that genetic and functional data point to a loss of enzyme function as the molecular mechanism of ARS-mediated recessive disease, impaired ARS activity likely leads to impaired protein translation. Since many ARS-associated recessive phenotypes include overlapping features, a global effect on protein translation may occur when any ARS enzyme is mutated; however, tissue-predominant effects may not be explained by a global reduction in protein translation.

One potential explanation for the observed tissue specificity of ARS-associated recessive disease is that tissues may have different demands for different amino acids (Figure 3A). For example, as discussed above, patients with CARS1 variants present with brittle hair and nails, which both express cysteine-rich keratins and keratin-associated proteins. Therefore, if tRNA^Cys charging with cysteine is impaired, hair and nails may be more susceptible to aminoacylation defects because of the high cysteine demand for keratin translation. Two lines of research would test this hypothesis. First, ribosome profiling studies [91] will be useful to determine if ribosomes are stalling at cysteine codons, in particular along transcripts with many cysteine codons. Second, quantitative proteomics studies using metabolic isotopic labeling [92] or label-free [93] quantitative mass spectrometry on relevant tissue samples would determine if the expression of cysteine-rich proteins is significantly reduced in the environment of pathogenic CARS1 variants.

An additional possibility to explain tissue-specific phenotypes associated with ARS variants relates to the availability of tRNA [59, 60] and amino acids [61, 62], which may vary among tissues—due to differences in pathways including biosynthesis and salvage—and influence the cellular response to impaired ARS activity (Figure 3A). For example, a mutant CARS1 enzyme with reduced activity would likely charge more tRNA^Cys molecules in a tissue with abundant tRNA^Cys and cysteine compared to a tissue with low availability of these molecules. Studies investigating tissue-specific tRNA expression and amino acid abundance are needed to address this hypothesis. First, to quantify tRNA abundance, tRNA sequencing techniques [94, 95] or tRNA microarrays [59] could be applied to different tissues, which would indicate if patient phenotypes occur in tissues with low levels of tRNA^Cys. Second, amino acid content across different tissues could be determined using chromatography [69, 96] to determine if patient phenotypes occur in tissues with low levels of cysteine.

A final consideration on the mechanisms behind tissue-predominant ARS-associated phenotypes is that secondary functions have been reported for some ARSs (Figure 3B). More basic science research is needed to uncover secondary functions of ARSs and to determine how these functions may relate to ARS-associated disease. Extending our CARS1 example reveals two interesting possibilities: (i) CARS1 functions in cysteine polysulfidation [97] and ferroptosis [98]; and (ii) alternative splicing results in two widely expressed CARS1 protein isoforms that contain different C-terminal amino-acid sequences [99]. Studies measuring cysteine hydropersulfide [100] in relevant patient tissues would indicate if cysteine polysulfidation is affected by impaired CARS1. Additionally, knockdown of CARS1 has been reported to inhibit ferroptosis induced by cysteine deprivation and to induce the transsulfuration pathway [98]; studies investigating components of these pathways in the environment of pathogenic CARS1 variants would be informative. To further investigate the two widely expressed CARS1 isoforms and their potential functions, subcellular fractionation [101] and fluorescence imaging [102] experiments would indicate if the isoforms have different subcellular localizations. Co-immunoprecipitation and mass spectrometry experiments [103] would also indicate if the isoforms have differential binding partners. By defining secondary functions of ARS enzymes, the community may then determine if—and how—pathogenic variants affect these activities.

Concluding Remarks

Variants in nearly all 37 human ARS loci have been implicated in genetic disease. Five ARS genes have been linked with dominant peripheral neuropathies, which may reflect sensitivity of the peripheral nervous system to impairments in protein translation and/or to aberrant protein interactions that affect signaling pathways. Variants in the majority of ARSs have been linked with recessive phenotypes and certain ARSs are associated with tissue-specific phenotypes despite the ubiquitous expression of these essential genes. The explanation for the observed tissue specificity is not understood and may reflect: (i) differing requirements across tissues to translate proteins enriched for specific amino acids; (ii) differing availability of tRNA and amino acids across tissues; and/or (iii) impairment of secondary functions of ARS enzymes that are important for specific tissues. Experiments designed to test these hypotheses will provide insights into the biological function of ARS enzymes, advance our understanding of the mechanisms of ARS-associated diseases, and more broadly enable us to characterize other clinical phenotypes caused by mutations in the protein translation machinery.

HIGHLIGHTS.

• Aminoacyl-tRNA synthetases (ARSs) are ubiquitously expressed, essential enzymes that ligate amino acids onto cognate tRNA molecules.

• Variants in loci encoding ARS enzymes have been implicated in myriad dominant and recessive human inherited disease phenotypes.

• Five of the 37 human ARS loci are associated with dominant peripheral neuropathy with a primarily axonal pathology; the vast majority of the alleles cause a loss-of-function effect.

• Nearly all 37 human ARS loci are associated with recessive multi-system disorders, and every genotype result in severely impaired ARS function.

• Interestingly, for both dominant and recessive phenotypes, certain tissues are exclusively or more severely affected; the mechanism(s) by which specific tissues are affected is not defined.

• Defining the molecular mechanism(s) of ARS-related disease and determining the cell-type specificity to impaired ARS function will provide insight into the biology of these essential enzymes and will reveal potential therapeutic targets.