|Home | About | Journals | Submit | Contact Us | Français|
Aminoacyl-tRNA synthetases (ARSs) are ubiquitously expressed enzymes responsible for ligating amino acids to cognate tRNA molecules. Mutations in four genes encoding an ARS have been implicated in inherited peripheral neuropathy with an axonal pathology, suggesting that all ARS genes are relevant candidates for disease in patients with related phenotypes. Here, we present results from a mutation screen of the histidyl-tRNA synthetase (HARS) gene in a large cohort of patients with peripheral neuropathy. These efforts revealed a rare missense variant (p.Arg137Gln) that resides at a highly conserved amino acid, represents a loss-of-function allele when evaluated in yeast complementation assays, and is toxic to neurons when expressed in a worm model. In addition to the patient with peripheral neuropathy, p.Arg137Gln HARS was detected in three individuals by genome-wide exome sequencing. These findings suggest that HARS is the fifth ARS locus associated with axonal peripheral neuropathy. Implications for identifying ARS alleles in human populations and assessing them for a role in neurodegenerative phenotypes are discussed.
Peripheral neuropathies are a heterogeneous class of disorders mainly characterized by distal motor and sensory dysfunction. Etiological classifications for peripheral neuropathies include acquired, syndromic, and inherited neuropathy, each of which can be further subdivided into axonal or demyelinating forms (Dyck and Lambert, 1968). The prevalence of peripheral neuropathy can exceed 10% in aging populations, making this class of diseases a major public health care concern in terms of proper diagnosis, prognosis, and therapeutic development (Martyn and Hughes, 1997). The prevalence and heterogeneity of peripheral neuropathies underscore the importance of identifying genetic lesions that directly cause or modify clinical phenotypes.
Aminoacyl-tRNA synthetases (ARSs) are ubiquitously expressed, essential enzymes that perform the first step of protein translation—attaching amino acids to cognate tRNA molecules (Antonellis and Green, 2008). Despite the ubiquitous function of ARSs, mutations in four loci encoding an ARS enzyme have been implicated in dominant and recessive peripheral neuropathies with an axonal pathology: glycyl-tRNA synthetase (GARS; MIM# 600287), tyrosyl-tRNA synthetase (YARS; MIM# 603623), alanyl-tRNA synthetase (AARS; MIM# 601065), and lysyl-tRNA synthetase (KARS; MIM# 601421) (Antonellis, et al., 2003; Jordanova, et al., 2006; Latour, et al., 2010; McLaughlin, et al., 2010). These data indicate that genes encoding ARS enzymes are relevant candidates for disease in patients with peripheral neuropathy and no known contributing genetic variation.
The full complement of ARS loci and alleles implicated in peripheral neuropathy remains undefined. Addressing this issue will aid in determining if peripheral nerve axons are susceptible to mutations in a specific subset of ARS enzymes or if mutations in any ARS enzyme will result in a peripheral neuropathy. Furthermore, the mechanistic link between ARS mutations and neurodegeneration has not been identified and many disease-associated alleles have not been evaluated for neurotoxicity in vivo. Establishing an approach for rapidly characterizing alleles in the 37 human ARS-encoding genes will be critical for advancing our understanding of ARS-related neurodegeneration. Here, we present data from patient-based mutation screening, yeast complementation assays, and a worm model of neurotoxicity that reveal a loss-of-function, neurotoxic histidyl-tRNA synthetase (HARS) variant. Our findings raise the possibility that HARS mutations are associated with axonal peripheral neuropathy and further support a role for impaired tRNA charging in the pathogenesis of ARS-related peripheral neuropathy.
DNA samples were isolated from 363 patients diagnosed with peripheral neuropathy by collaborating physicians. DNA samples from the ClinSeq™ cohort were collected as previously described (Biesecker, et al., 2009). The appropriate review boards of each participating institution approved these studies. The HARS gene was assessed for mutations at the NIH Intramural Sequencing Center (NISC) as previously described (McLaughlin, et al., 2010). Each HARS variant was independently confirmed via PCR amplification followed by DNA sequencing analysis. Nucleotide positions are relative to the open reading frame from GenBank Accession No. NM_002109.3. Nucleotide numbering reflects cDNA position with +1 corresponding to the A of the ATG translation initiation codon in the reference sequence. Codon numbering for the HARS protein was based on the reported amino acid sequence (NCBI accession no. NP_002100.2). The initiation codon is codon 1. The variants were submitted to an LSDB at www.LOVD.nl/HARS.
A total of 576 DNA samples isolated from neurologically normal controls were obtained from the National Institute of Neurological Disease (NINDS) and Coriell Institute for Medical Research (Panel numbers NDPT079, NDPT082, NDPT084, NDPT090, NDPT093, and NDPT094) and genotyped for each HARS variant using MassARRAY® iPLEX Gold technology (Sequenom, San Diego, CA). Briefly, multiplex HARS-specific PCR reactions were performed (primer sequences available upon request) followed by primer extension assays using mass-modified dideoxynucleotide terminators. MALDI-TOF mass spectrometry was used to determine the genotype for each individual using SpectroTYPER software (Sequenom). The presence of each HARS variant in the general population was additionally assessed using the Exome Variant Server (snp.gs.washington.edu/EVS/), the 1000genomes database (www.1000genomes.org/), and dbSNP (www.ncbi.nlm.nih.gov/snp).
HARS protein orthologs from multiple species were derived from the following GenBank accession numbers: Human (Homo sapiens, NP_002100.2), Mouse (Mus musculus, NP_032240.3), Chicken (Gallus gallus, NP_001006144.1), Zebrafish (Danio rerio, NP_001004586.1), Nematode (Caenorhabditis elegans, NP_001023373.1), Fruit Fly (Drosophila melanogaster, NP_728180.1), Baker’s Yeast (Saccharomyces cerevisiae, AAA34696.1), and Bacteria (Escherichia coli, NP_289067.1). Amino acid sequences were aligned using Clustal W2 (http://www.ebi.ac.uk/Tools/clustalw2/index.html).
Each HARS protein variant was analyzed using the MUPro (http://www.ics.uci.edu/~baldig/mutation.html), PolyPhen (http://genetics.bwh.harvard.edu/pph/index.html), PolyPhen2 (http://genetics.bwh.harvard.edu/pph2/), and SIFT (http://sift.jcvi.org/www/SIFT_seq_submit2.html) prediction algorithms.
DNA constructs for yeast complementation assays were generated using Gateway cloning technology (Invitrogen, Carlsbad, CA). The S. cerevisiae HTS1 locus (including the open reading frame and 477 base pairs of proximal promoter sequence) were PCR amplified using S. cerevisiae genomic DNA. Each primer was designed to include flanking Gateway attB1 (forward) and attB2 (reverse) sequences (primers available upon request). Subsequently, purified PCR products were recombined into the pDONR221 vector (according to the manufacturer’s instructions) to yield an appropriate entry clone; the insert of the final entry clone was sequenced to confirm specificity. Mutagenesis of HTS1 was performed using the QuikChange II XL Site-Directed Mutagenesis Kit (Stratagene, Santa Clara, CA) and appropriate mutation-bearing oligonucleotides. Subsequently, plasmid DNA purified from each entry clone was recombined with a gateway-compatible pRS315 or pRS316 destination vector according to the manufacturer’s instructions. Each resulting expression construct was analyzed by restriction enzyme digestion with BsrGI (New England Biosystems, Ipswich, MA) to confirm the presence of an appropriately sized insert.
A commercially available diploid heterozygous hts1Δ yeast strain (MATa/α, his3Δ1/his3Δ1, leu2Δ0/leu2Δ0, LYS2/lys2Δ0, met15Δ0/MET15, ura3Δ0/ura3Δ0; Open Biosystems, Hunstville, AL) was transformed with a URA3-bearing pRS316 vector containing wild-type HTS1 (see above). Lithium acetate transformations were performed at 30°C using 200 ng of plasmid DNA, as previously described (Green, et al., 1999). Transformation reactions were grown on yeast medium lacking uracil. Subsequently, sporulation and tetrad dissections were performed to obtain a haploid hts1Δ strain that also contains the wild-type HTS1 pRS316 maintenance vector. Briefly, the transformed diploid hts1Δ strain was patched twice onto presporulation plates (5% D-glucose [Fisher Scientific, Hampton, NH], 3% nutrient broth [Becton, Dickinson and Company, Franklin Lakes, NJ], 1% yeast extract [Acros Organics, Waltham, MA], and 2% agar [Teknova, Hollister, CA]). One microliter of cells was then transferred into 2 ml of supplemented liquid sporulation medium (1% potassium acetate [Fisher Scientific], 0.005% zinc acetate [Fisher Scientific], 1X ura supplement [MP Biomedicals, Solon, OH], 1X his supplement [MP Biomedicals], and 1X leu supplement [MP Biomedicals]) and incubated for 5 days at 25°C followed by 3 days at 30°C. Sporulated strains were dissected using a MSM 400 dissection microscope (Singer Instruments, Somerset, UK), and plated on yeast extract, peptone, and dextrose (YPD) plates (Becton, Dickinson and Company). Resulting spores were individually patched onto solid growth medium containing Geneticin (G418) or 0.1% 5-fluoroorotic acid (5-FOA), or media lacking uracil (Teknova). Two spores that grew on G418 and yeast medium lacking uracil, yet did not grow on 5-FOA medium, were selected for use in the yeast viability assays. Two resulting haploid hts1Δ strains bearing a wild-type HTS1 pRS316 maintenance vector were then transformed with LEU2-harboring pRS315 constructs containing wild-type or mutant HTS1 (described above) and grown on medium lacking uracil and leucine (Teknova). For each transformation, five colonies were selected for additional analysis. Each colony was diluted in 100 μl H2O, then further diluted 1:10 and 1:100, and spotted on growth medium containing 0.1% 5-FOA (Teknova), incubated for 3 days at 30°C, and growth was assessed by visual inspection. Because 5-FOA is toxic to yeast cells bearing a functional URA3 allele, only cells that lost the URA3-bearing maintenance plasmid and for which the LEU2-harboring test plasmid could complement the chromosomal hts1Δ allele were expected to grow. To generate growth curves, surviving yeast colonies from strains harboring p.Val238Ala and p.Pro505Ser HTS1 were inoculated in selective medium and incubated shaking overnight at 30°C. Dilutions were generated in YPD medium to obtain an OD660 of 0.05, with the new cultures incubated shaking at 30°C. The OD660 was then measured at 1, 2, 3, 4, 5, 6, 7, and 25 hours using a NanoDrop 1000 (Thermo Fisher Scientific, Waltham, MA). The percent increase in growth was determined by dividing each of the measured OD660 values by the initial OD660 value and expressing relative to the wild-type HTS1-harboring strain.
Standard methods for maintaining C. elegans were used (Brenner, 1974). Bristol strain N2 was used as the wild type, and all animals were grown at 20°C. All C. elegans strains were obtained from the C. elegans Genetics Stock Center, (University of Minnesota).
C. elegans transgene constructs were made from the expression vector pPD157.60 (Addgene, Firelab vector kit), which contains promoter sequences from the C. elegans glutamic acid decarboxylase (unc-25) gene. Expression of C. elegans wild-type or mutant hars-1 cDNAs were engineered by cloning sequence verified full-length wild type or mutant R137Q cDNA sequences into the XbaI/NaeI sites of pPD157.60. Green fluorescent protein (GFP) fusion constructs were engineered using a PCR fusion approach to generate in-frame N-terminal tagged Punc-25::GFP-hars-1 (WT or R137Q) plasmids (Hobert, 2002).
Transgenic C. elegans strains were generated by standard microinjection techniques (Mello, et al., 1991). All transgenic strains were injected with plasmids of interest at 50 ng/ml + 2.5 ng/ml pCFJ90 (Pmyo-2::mCherry) + 47.5 ng/ml 1 Kb+ DNA Ladder (Invitrogen). For all strains, stable transgenic lines were selected based on mCherry expression in the pharyngeal muscles from a Pmyo-2:mcherry coinjection marker. At least three independent lines were isolated and analyzed. The data shown are from one representative line.
Movement defects were quantified by counting the number of body thrashes in liquid for a two-minute interval (Miller, et al., 1996). Specifically, 100 microliters of room temperature M9 media was aliquoted into an agar-coated 96 well plate. Individual animals were picked to wells and allowed to acclimate for 2 minutes. Movement was recorded using a cooled monochrome camera (Leica) and the number of times animals bent across their body axis was counted. To allow for accurate scoring of body bends, the image sequence was played at one-quarter speed offline using ImageJ. The total number of body bends counted was divided by two, yielding a body thrash/minute count.
Motor neuron quantification was performed on the following strains: EG1285: oxIs12 [Punc-47::GFP; lin-15b] X, BEG9: oxIs12 [Punc-47::GFP, lin-15b] X; aabEx8 [Punc-25::hars-1 (WT), Pmyo-3::mCherry], BEG10 : oxIs12 [Punc-47::GFP, lin-15b] X; aabEx9 [Punc-25::hars-1 (R137Q), Pmyo-3::mCherry], BEG11: aabEx10 [Punc-25::GFP-hars-1 (WT), Pmyo-3::mCherry], BEG12: aabEx11 [Punc-25::GFP-hars-1 (R137Q), Pmyo-3::mCherry. Synchronized fourth larval stage worms were grown at 20°C. L4, one-day and four-day adults were selected for imaging and phenotypic characterization. To visualize axonal processes, transgenic worms were anesthetized in M9 buffer containing 5 mM levamisole, mounted on agar pads and then coverslipped. The quantify the percent animals exhibiting aberrant GABA motor neurons commissures, >100 animals/genotype/timepoint were imaged. Animals exhibiting at least one aberrant commissure were scored as positive.
Animals were imaged on a Leica DMI6000B compound microscope with a CCD camera (FDC360, Leica) using a 40X objective. Confocal images were obtained on a Nikon A1R microscope with a 20X objective.
We screened the histidyl-tRNA synthetase (HARS) gene for protein-coding variations in 363 patients with peripheral neuropathy and no previously identified mutation (the appropriate, institute-specific review boards approved all studies performed herein, and informed consent was obtained from all subjects). These efforts revealed six HARS missense variants in the heterozygous state in eight patients from our cohort: c.14C>A which predicts p.Ala5Glu, c.410G>A which predicts p.Arg137Gln, c.605G>A which predicts p.Gly205Asp, c.713T>C which predicts p.Val238Ala, c.1127A>G which predicts p.Lys376Arg, and c.1513C>T which predicts p.Pro505Ser (Table 1).
To determine the frequency of each identified HARS variant we tested for the presence of each in five available resources: (1) dbSNP (locus identifier 3035); (2) The NINDS/Coriell neurologically normal control panels (NDPT079, NDPT082, NDPT084, NDPT090, NDPT093, and NDPT094); (3) The Exome Variant Server (EVS; snp.gs.washington.edu/EVS/); (4) The 1000genomes database (browser.1000genomes.org); and (5) The ClinSeq™ patient cohort (Biesecker, et al., 2009). This revealed the presence of p.Ala5Glu, p.Arg137Gln, p.Gly205Asp, and p.Lys376Arg HARS in one or more of these analyses, while p.Val238Ala and p.Pro505Ser HARS were not detected in any of the five resources (Table 1). We also collected DNA samples from additional affected and unaffected family members related to the two carriers of p.Gly205Asp HARS and the single carrier of p.Lys376Arg HARS. Neither of these variants segregate with disease status in the three pedigrees (data not shown). Additional family members were not available to test for segregation of the remaining four variants with disease status.
To predict the pathogenic potential of the identified HARS variants we assessed the evolutionary conservation of each affected amino acid. Briefly, we collected the protein sequences of HARS orthologs from multiple, evolutionarily-diverse species, aligned the sequences with ClustalW software (Larkin, et al., 2007), and visually inspected each affected amino acid (Fig. 1B). Alanine 5 resides in a region of the protein sequence that is difficult to align, and is conserved among three of the five species analyzed. Arginine 137 is conserved among all species analyzed, including bacteria—indeed, this amino-acid residue is invariant among all available protein sequences (Freist, et al., 1999). Glycine 205 is conserved among six of the eight species analyzed, including worm and fruit fly. Valine 238 is conserved among seven of the eight species analyzed, including yeast. Lysine 376 is conserved among four of the eight species analyzed, all of which are vertebrate species. Finally, proline 505 also resides in a region of the protein sequence that is difficult to align and is conserved among two of the eight species analyzed.
We applied computational predictions to further assess the potential effect of each HARS variant on protein function using the MuPro, PolyPhen, PolyPhen2, and SIFT algorithms (Cheng, et al., 2006) (Table 2), which have been employed to assess disease-associated aminoacyl-tRNA synthetase mutations (McLaughlin, et al., 2010). The HARS variants p.Ala5Glu and p.Lys376Arg were not predicted to be pathogenic by any of the algorithms, and p.Gly205Asp was only predicted to be pathogenic by one of the four algorithms. In contrast, p.Val238Ala and p.Pro505Ser were predicted to be pathogenic by three of the four algorithms, and p.Arg137Gln was predicted to be pathogenic by all four algorithms employed.
We next mapped each variant onto the primary structure of the HARS protein (Fig. 1C). Alanine 5 is located at the first amino acid of the WHEP-TRS domain, which is predicted to mediate the association of HARS with the multiple aminoacyl-tRNA synthetase complex (pfam00458 in the NCBI Conserved Domain Database). Interestingly, glycyl-tRNA synthetase (GARS) has a similar N-terminal domain (Antonellis, et al., 2003) and, to date, no disease-associated mutations have been identified therein. Arginine 137, glycine 205, valine 238, and lysine 376 reside within the catalytic core of the enzyme and proline 505 resides within the tRNA-binding domain (Fig. 1C). Thus, the variants that affect these amino acids may be detrimental to enzyme kinetics or tRNA recognition. Importantly, previously described disease-associated ARS mutations impair these two processes (Antonellis and Green, 2008; Antonellis, et al., 2006; Latour, et al., 2010; McLaughlin, et al., 2012; McLaughlin, et al., 2010; Nangle, et al., 2007).
Yeast complementation assays have been successfully employed to evaluate the functional consequences of GARS, YARS, AARS, and KARS mutations. (Antonellis, et al., 2006; Jordanova, et al., 2006; McLaughlin, et al., 2012; McLaughlin, et al., 2010). Importantly, each ARS allele that is unable to complement deletion of the yeast ortholog is also unable to charge tRNA molecules as assessed by in vitro aminoacylation assays indicating that yeast is a relevant in vivo model to test for a loss-of-function effect (Antonellis, et al., 2006; Jordanova, et al., 2006; Nangle, et al., 2007; McLaughlin, et al., 2012; McLaughlin, et al., 2010). To experimentally evaluate p.Arg137Gln, p.Gly205Asp, p.Val238Ala, p.Lys376Arg, and p.Pro505Ser HARS for a loss-of-function effect in vivo, we modeled the missense changes in the yeast ortholog HTS1 (Table 3) and tested the ability of each variant to support yeast cell growth in complementation assays—p.Ala5Glu HARS was not modeled due to the inability to predict the orthologous residue in yeast (Fig. 1B). Briefly, a haploid yeast strain was generated with the endogenous HTS1 gene deleted (hts1Δ) and viability supported via a maintenance vector consisting of a wild-type copy of HTS1 on a URA3-bearing plasmid (pRS316). Experimental alleles were generated on a LEU2-bearing vector (pRS315), transformed into the above strain, and viability assessed by analysis of growth on 5-fluoroorotic acid (5-FOA), which selects for spontaneous loss of the URA3-bearing maintenance vector (Boeke, et al., 1984). The wild-type HTS1 vector supported significant growth, while an insert-free pRS315 construct did not (Fig. 2A), consistent with our experimental vector harboring a functional copy of HTS1, and with HTS1 being an essential gene, respectively. The p.Gly205Asp, p.Val238Ala, p.Lys376Arg, and p.Pro505Ser HTS1 variants allowed growth in a manner consistent with wild-type HTS1. Since p.Val238Ala and p.Pro505Ser HARS were not detected in the public variant databases (Table 1), we further explored them for a loss of function via growth curve analysis. Neither of these variants grew in a manner significantly different from wild-type HTS1 over a 25-hour period (Fig. 2B). In contrast to the above results, p.Arg137Gln HTS1 could not complement the hts1Δ allele (Fig. 2A). Combined, our functional and computational analyses strongly suggest that p.Arg137Gln represents a loss-of-function allele and that it is the most likely of the HARS variants to be pathogenic. This notion is based on observations that previously identified disease-associated aminoacyl-tRNA synthetase mutations occur at highly conserved amino acids, are computationally predicted to be pathogenic, and cause a loss of function in aminoacylation and yeast growth assays (Antonellis, et al., 2003; Antonellis, et al., 2006; Jordanova, et al., 2006; Latour, et al., 2010; McLaughlin, et al., 2012; McLaughlin, et al., 2010; Nangle, et al., 2007).
The p.Arg137Gln HARS variant was identified in a patient with sporadic motor and sensory peripheral neuropathy, with sensory deficits more severe than motor deficits. At examination, the patient was a 64-year-old male with a 15-year history of impaired sensation in the lower extremities. There was no history of painless ulcers and he suffered from shooting pains in his legs over the previous four years. Pes cavus and hammer toes were absent and there was no overt muscle wasting. There were brisk knee jerks, absent ankle jerks and the plantar reflexes were flexor. Previous nerve conduction studies at age 57 showed a mild slowing of peroneal motor conduction velocity at 37 m/sec with a distal latency of 6.6 m/sec and an amplitude of 5.60 microvolts, and a sural sensory amplitude of 6.5 uV. At age 64 there were absent sural responses and a borderline reduced (10uV) and delayed (3.82 m/secs) median sensory action potential. The peroneal motor action potential was not obtained and the median motor conduction velocity was 50 m/sec. Combined, these data indicate a neuropathy of axonal (or mixed axonal and demyelinating) origin. Interestingly, neurological examination of the 57 year-old female ClinSeq™ subject who is a carrier of p.Arg137Gln HARS (Table 1) did not reveal peripheral neuropathy—Sanger sequencing confirmed that this individual indeed carries this variant (data not shown). Given the late age of disease onset in the original patient and the incomplete penetrance of certain neuropathy-associated ARS mutations (Dubourg, et al., 2006; Sivakumar, et al., 2005), determining the pathogenic potential of p.Arg137Gln HARS in a relevant animal model system was considered essential for confirming the pathogenicity.
To evaluate the neurotoxicity of p.Arg137Gln HARS, we used the genetically tractable soil nematode, Caenorhabditis elegans, as an experimental model. As in humans, the C. elegans genome contains a single gene encoding the cytoplasmic histidyl-tRNA synthetase (hars-1). The C. elegans HARS-1 protein is 54% identical and 70% similar to the human ortholog HARS (data not shown). Due to the conservation of arginine 137 between human and worm (Fig. 1B), we hypothesized that C. elegans hars-1 might be used as a template to screen and differentiate mutations that impart neuronal toxicity in vivo. To this end, we over-expressed C. elegans hars-1 variants specifically in GABAergic neurons, most of which are subclasses of motor neurons in nematode, to determine if p.Arg137Gln is sufficient to cause functional motor neuron defects. To engineer the GABA motor neuron specific expression of HARS-1 in C. elegans, a chimeric transgene was constructed in which the promoter region of the C. elegans unc-25 gene, which encodes for glutamic acid decarboxylase, directs the expression of a full-length wild-type or p.Arg137Gln hars-1 cDNA [Punc-25::hars-1(WT) or Punc-25::hars-1(p.Arg137Gln)] (Schuske, et al., 2004). To assay motor neuron abnormalities associated with transgene expression, the Punc-25::hars-1 (WT or p.Arg137Gln) transgene was introduced into an integrated strain containing the oxIs12 reporter transgene (McIntire, et al., 1997), which expresses GFP under the control of the GABA motor neuron specific unc-47 promoter [Punc-47::GFP] (Fig. 3A and 3B). This strain allows for the visualization of all GABA motor neuron cell bodies and their fine axonal processes (Fig. 3B). Motor neuron specific expression of p.Arg137Gln, but not WT, induced significant morphological abnormalities in GABA motor neurons (Fig. 3C and 3D). Expression of the p.Arg137Gln variant resulted in aberrant commissural axonal processes in ~40% of animals analyzed at three different developmental stages (L4, 1-day adult and 4-day adult; Fig. 3G). Specifically, animals bearing the p.Arg137Gln transgene exhibited gross morphological defects in commissural axons, denoted by failure to reach the dorsal nerve cord, axonal beading, defasiculation, and breaks in the visualized GABAergic dorsal nerve cord (Fig. 3D). None of these defects were observed in transgenic control animals or animals over-expressing a wild-type transgene (Fig. 3B and 3C).
To determine if the morphological aberrations observed in the p.Arg137Gln transgenic animals were due to mislocalization of the HARS-1 protein, we constructed wild-type and p.Arg137Gln GFP fusion proteins directed by the unc-25 promoter [Punc-25::GFP-hars-1(WT) or Punc-25::GFP-hars-1(p.Arg137Gln)]. Heritable transgenic strains were generated by microinjection into N2 wild-type C. elegans strains (Mello and Fire, 1995; Mello, et al., 1991). Both the wild-type and p.Arg137Gln GFP fusion proteins localized specifically to GABA motor neurons and GFP was observed in the cell bodies and axonal processes (Fig. 3E and 3F). Importantly, the GFP-p.Arg137Gln fusion protein was able to recapitulate many of the axonal defects observed in transgenic animals over-expressing the non-tagged p.Arg137Gln variant (Fig. 3F)—these defects were not seen in animals expressing the wild-type HARS-1 GFP fusion protein (Fig. 3E).
Lastly, we tested whether GABA neuron-specific expression of hars-1 transgenes could induce motor performance defects. In addition to aberrant commissural axonal processes, p.Arg137Gln transgenic animals displayed mild locomotor defects on agar plates, whereas wild-type expressing strains appeared normal (data not shown). These defects were more severe in liquid media where thrashing rates (Miller, et al., 1996) of p.Arg137Gln transgenics were significantly lower than for the control or wild-type expressing strains (Fig. 3H). Interestingly, the aberrant axonal phenotype was present at L4 through 4-day adults, but the functional defect only manifests in 4-day adults (Fig. 3H). These data suggest that the p.Arg137Gln mutant exhibits a progressive loss of motor neuron function and coordination. Taken together, the in vivo morphological and functional defects suggest that p.Arg137Gln HARS imparts a cell autonomous neurotoxicity to motor neurons.
Here, we describe mutation analysis of the histidyl-tRNA synthetase (HARS) gene in patients with peripheral neuropathy. Six missense variants were identified and characterized, and computational and functional analyses revealed the following features of the p.Arg137Gln HARS variant: (1) Location within the catalytic domain of the HARS protein and at an amino-acid residue conserved between human and bacteria; (2) Loss-of-function based on yeast complementation assays; and (3) Dominant toxicity to neurons in a C. elegans model. Importantly, we have recently validated the C. elegans model of neurotoxicity via analysis of the most common neuropathy-associated ARS mutation, R329H alanyl-tRNA synthetase [(AARS); personal communication, A.A. Beg, manuscript in preparation]. Combined, these data indicate that p.Arg137Gln HARS may be a pathogenic allele similar to other disease-associated aminoacyl-tRNA synthetase mutations—such mutations are found in patients with motor and sensory peripheral neuropathy and an axonal pathology, cause a loss of function in tRNA charging, yeast viability, or proper localization within axons, and are dominantly toxic to axons in animal model systems (Antonellis, et al., 2003; Antonellis, et al., 2006; Jordanova, et al., 2006; Latour, et al., 2010; McLaughlin, et al., 2012; McLaughlin, et al., 2010; Nangle, et al., 2007; Seburn, et al., 2006; Storkebaum, et al., 2009). Furthermore, arginine 137 resides in the BM12 domain of the HARS protein, which is involved in enzyme dimerization (Freist, et al., 1999)—the majority of disease-associated GARS mutations also affect amino acids that reside on the dimer interface (Cader, et al., 2007; He, et al., 2011; Nangle, et al., 2007; Xie, et al., 2007). Our findings expand the panel of loss-of-function human ARS alleles, and the data from our worm experiments indicate that p.Arg137Gln HARS is toxic to neurons in vivo.
The mechanistic link between heterozygosity for loss-of-function ARS alleles and axonal degeneration remains unknown. Haploinsufficiency of tRNA charging seems an unlikely explanation based on data showing that most disease-associated mutations are missense changes and that heterozygosity for a Gars null allele does not cause neuropathy in mouse (Seburn, et al., 2006). One possibility is that mutant ARS monomers interfere with the activity of the wild-type protein by a dominant-negative effect. This is consistent with the observation that all ARS enzymes implicated in axonal degeneration to date act as homodimers—HARS also functions as a homodimer (Freist, et al., 1999) and p.Arg137Gln does not affect enzyme dimerization (data not shown). This possibility is further supported by the identification of compound heterozygosity for a null allele and a hypomorphic allele at the lysyl-tRNA synthetase (KARS) locus in a patient with peripheral neuropathy (McLaughlin, et al., 2010). The combination of these alleles is predicted to cause a depletion of KARS activity below 50% and to possibly cross a threshold of tRNA charging required for axon function. However, over-expression of wild-type human GARS in two mouse models of dominant peripheral neuropathy caused by heterozygosity for Gars mutations did not dramatically improve the phenotype (Motley, et al., 2011), indicating that any dominant-negative effect may be irreversible. However, it will be important to determine if wild-type HARS-1 can complement the toxicity of p.Arg137Gln HARS-1 in a stable transgenic C. elegans model system using Mos1-mediated transgenesis. Another possibility is that impaired tRNA charging or impaired ARS localization are prerequisites for some undefined gain-of-function toxicity to axons (e.g., these RNA-binding proteins may now be free to inappropriately interact with RNA molecules in the axon or nerve cell body). Finally, the neurotoxic effect of ARS mutations is unlikely to involve mitochondrial enzyme function. While GARS and KARS charge tRNA in both the cytoplasm and mitochondria, YARS, AARS, and HARS are predicted to charge tRNA only in the cytoplasm. Furthermore, disease-associated mutations in the genes encoding mitochondrial ARS enzymes (AARS2, DARS2, EARS2, HARS2, MARS2, RARS2, SARS2, and YARS2) are not associated with an overt axonal neuropathy (Bayat, et al. 2012; Belostotsky, et al. 2011; Edvardson, et al., 2007; Gotz, et al. 2011; Pierce, et al. 2011; Riley, et al. 2010; Scheper, et al., 2007; Steenweg, et al. 2012); however, detailed neurological examinations of patients with mitochondrial ARS mutations will be required to rule out such a phenotype.
Inherited peripheral neuropathies are clinically and genetically heterogenous with variable age of onset and reduced penetrance associated with specific loci. Moreover, the extent to which genetic (i.e., due to inherited or de novo mutations) versus environmental causes (e.g., diabetes and vitamin deficiencies) account for the 10% prevalence of peripheral neuropathy in the aging population (Martyn and Hughes, 1997) is unknown. As such, these diseases present a major challenge for the identification and validation of disease-causing or disease-modifying variants, especially when faced with small pedigrees or sporadic cases with one or both biological parents unavailable to evaluate for de novo mutations, thus hindering the ability to build a robust genetic argument. Indeed, the identification of p.Arg137Gln HARS in one affected and one unaffected individual underscores the importance of informative functional analyses. The studies outlined here have two major implications for the above challenges. First, we present a pipeline to rapidly assess the loss of function and axonal toxicity of ARS alleles in vivo. This approach will aid in the distinction between pathogenic and non-pathogenic ARS alleles and is feasible for variants identified at all 37 human ARS loci (Antonellis and Green, 2008). Second, these findings underscore the importance of identifying ARS alleles with neurotoxic potential in patient cohorts or in the general population. Alleles such as p.Arg137Gln HARS may directly predispose patients to peripheral neuropathy or may modify a peripheral neuropathy phenotype by contributing to the genetic and environmental load (Lupski, et al., 2011) in a given patient. However, while our functional analyses revealed a loss-of-function neurotoxic HARS allele, the causal link between HARS mutations and peripheral neuropathy remains unclear, largely due to the lack of a strong genetic argument. Studies designed to further screen relevant patient populations for HARS mutations, clinically evaluate other carriers of p.Arg137Gln HARS, and develop a mammalian model of p.Arg137Gln HARS will be critical for addressing this issue. In summary, our approaches and findings have important implications for understanding aminoacyl-tRNA synthetase variants and how they relate to human disease as well as a basis for understanding such variation in the context of gene X gene and gene X environment interactions.
This work was supported in part by grants NS060983 (AA), R01NS052767 (SZ), and U54NS065712 (SZ) from the National Institute of Neurological Diseases and Stroke (AA) and by the Intramural Research Programs of the National Human Genome Research Institute and National Institute of Neurological Diseases and Stroke (National Institutes of Health).
We are indebted to the patients and their families for participating in this study. We thank Ellen Pederson, Bob Lyons, and the University of Michigan DNA Sequencing Core for sequencing and genotyping analysis; Jennifer Johnston for mutation validation; Marina Kennerson for additional patient information; Chani Hodonsky for expert technical assistance; Tom Wilson for instruction on yeast viability assays; Eric Green for ARS locus sequencing; Sundeep Kalantry for critical analysis of the manuscript; and Miriam Meisler and her laboratory for thoughtful discussions on the experiments and manuscript. Some nematode strains were provided by the Caenorhabditis Genetics Center.