Therapeutic rescue of pathogenic asparaginyl-tRNA synthetase alleles

Abstract

Pathogenic alleles in the cytoplasmic asparaginyl-tRNA synthetase (NARS1) are associated with infant- and juvenile-onset disease, with no current disease-specific treatments. We developed a tractable human cell system to study disease-causing NARS1 alleles that can be adapted to investigate NARS1 and other aminoacyl-tRNA synthetase (ARS) alleles. We found that two dominant NARS1 nonsense alleles, R534X and R522X, cause a cytotoxic phenotype and elicit the integrated stress response (ISR). Proteomic and phenotypic changes were rescued by asparagine supplementation in the human cell model. Asparagine supplementation completely restored cell proliferation defects in patient-derived fibroblasts and prevented activation of the ISR. We also tested therapeutic cognate transfer RNA (tRNA) supplementation, which reduced the cytotoxicity of pathogenic NARS1 alleles but did not ameliorate activation of the ISR. A general control nonderepressible 2 (GCN2) inhibitor suppressed ISR activation and reduced cytotoxicity but did not restore changes to the proteome caused by the NARS1 nonsense alleles. The data reveal molecular and cellular defects caused by premature termination codons in NARS1 alleles. Our data also indicate asparagine supplementation as a feasible therapeutic approach to address the underlying cause of NARS1 disease, a rare disease for which currently no treatment is available.

Graphical abstract

Heinemann and colleagues show that pathogenic NARS1 nonsense variants trigger cellular stress and impaired growth and that asparagine supplementation reverses these defects in human cells and patient fibroblasts. The findings identify a promising therapeutic strategy for NARS1-associated disease and demonstrate a scalable platform for studying aminoacyl-tRNA synthetase disorders.

Introduction

Cytoplasmic asparaginyl-tRNA synthetase (NARS1) disease is an inherited disorder with broad effects on the central nervous system (CNS) and peripheral nervous system (PNS). The prevalence of NARS1 disease is extremely low, with 54 cases across 30 distinct families currently reported.¹ NARS1 disease can be inherited in an autosomal-dominant (AD), autosomal-recessive (AR), or de novo fashion.² Generally, dominant NARS1 alleles cause Charcot-Marie-Tooth (CMT) disease, while autosomal-recessive variants are typically associated with developmental delay and microcephaly. Bi-allelic and homozygous variants present distinct neuropathological symptoms in NARS1 patients, with the homozygous alleles being subcategorized as NARS1-related hereditary neuropathy. This subtype manifests in adolescence and early adulthood with mostly motor and sensory impairment involving weakness of predominantly the distal lower limbs and foot deformities without prominent muscle atrophy.^1,3

Interestingly, dominant heterozygous nonsense alleles R522X and R534X causing premature protein termination in the C-terminal domain result in a non-CMT phenotype, NARS1-related neurodevelopmental disorder (NARS1-NDD), which presents neonatal and early-childhood symptoms including global developmental delay, intellectual disability, microcephaly, ataxia, seizures, and, rarely, neurobehavioral/psychiatric manifestations.¹ These individuals also possess difficulty coordinating movements, speech problems, and a decline in intellectual function.⁴ Currently, no disease-specific treatment exists, and symptoms are managed with supportive measures and surgeries.¹

NARS1 is a member of a family of highly conserved, essential enzymes that play a vital role in protein synthesis, known as aminoacyl-tRNA synthetases (ARSs). In eukaryotes, each of the 20 proteinogenic amino acids has its own specific ARS enzyme⁵ that ligates amino acids to their cognate transfer RNAs (tRNAs). A high degree of selectivity in the ARS active site is critical, as the ribosome cannot distinguish a correctly charged from a mischarged tRNA. Disease-causing mutations in ARS proteins can lead to a multitude of molecular and cellular consequences,^6,7,8 including decreased ARS stability, deficient aminoacylation activity,^9,10,11 or toxic gain of function, since some alleles in the catalytic site result in mis-aminoacylation and proteome-wide mistranslation.¹² Disease-causing mechanisms that are independent of aminoacylation have also been described.^13,14

In humans, the cytoplasmic (NARS1) and mitochondrial (NARS2) NARS proteins are encoded by distinct genes.¹⁵ Several single-nucleotide polymorphisms (SNPs) in the NARS1 gene have been identified as disease causing.^2,4,16 Two early-onset cases are caused by nonsense mutations of arginine (R, Arg) codons (R522X and R534X), leading to premature termination codons in NARS1 transcripts associated with NARS1-NDD. These NARS1 Arg residues (R522 and R534) are conserved across eukaryotes and are localized in the C-terminal catalytic domain of the protein, which contains the active site² (Figure 1A). A recent study characterized the R534X recombinant protein in detail and found that its dominant effect is based on preferential formation of heterodimers that are catalytically incompetent.^17,18 The molecular and cellular consequences of these alleles remain largely uncharacterized.

Figure 1.

Pathogenic allele localization in human asparaginyl-tRNA synthetase and cellular consequences

(A) The crystal structure of human NARS1 lacking the UNE-E was visualized in visual molecular dynamics (VMD) (PDB: 8TC8) with one monomer depicted in tan and one in teal. C-terminal peptides that are truncated in R522X and further amino acids that are C-terminal to R534X are depicted in purple and orange. (B) Molecular and cellular consequences of pathogenic NARS alleles in the cell. Dominant-negative NARS alleles lead to reduced aminoacylation and the accumulation of uncharged tRNA, which activates the integrated stress response.

Several ARS mutations, including those found in the alanyl-tRNA synthetase (AARS1) editing active sites^19,20 and disease-causing variants,²¹ are known to induce the integrated stress response (ISR).²² The ISR is stimulated by stress-responsive kinases that are activated by disruptions in protein homeostasis, including uncharged tRNA accumulation (Figure 1B) due to amino acid starvation.^23,24 Uncharged tRNAs are recognized by the general control nonderepressible 2 (GCN2) kinase, which then phosphorylates eukaryotic initiation factor 2α (eIF2α) at serine 51.²⁴ Phospho-eIF2α inhibits eukaryotic translation initiation factor 2B (eIF2B), preventing guanine nucleotide exchange activity that is required for the delivery of the initiator tRNA to the ribosome at the start of protein synthesis. As a result, global translation is attenuated and activating transcription factor 4 (ATF4) is expressed, favoring the subsequent expression of genes involved in cellular adaptation and amino acid synthesis.²⁴ ISR activation is thought to contribute to disease phenotypes and is activated in glycyl-tRNA synthetase (GARS1) mouse models of CMT disease.²² Pharmacological inhibition of GCN2 is efficacious in improving a GARS1 mouse model’s neuropathological phenotype, highlighting the potential of this treatment approach.²²

Here, we investigated the molecular and cellular consequences of two pathogenic NARS1 alleles, R522X and R534X, to test the hypothesis that NARS1 disease-associated mutations reduce aminoacylation and cause proteomic perturbations that can be rescued by supplementation with asparagine (Asn), cognate tRNA^Asn, or GCN2 inhibitors. We found that the nonsense alleles elicit a cytotoxic phenotype and activate the ISR in an HEK293T cell model, which was then rescued using Asn or a GCN2 inhibitor. Using mass spectrometry (MS), we identified proteomic perturbations from overexpression of the R522X mutant, with pathways like the ISR, and vascular endothelial growth factor signaling-associated proteins returning to wild-type (WT) levels when treated with Asn. Further, we used patient-derived primary fibroblasts from an individual with the NARS1 R534X variant and found that decreased cell viability is rescued with Asn supplementation.

Results

Recombinant NARS1 variants are thermally stable but prone to degradation

Pathogenic alleles in several ARSs were shown to elicit changes in protein structure and stability.^{9,11,12,25,26,27} To assess whether the NARS1 variants result in changes to the thermal stability or protein folding, we performed differential scanning fluorimetry on recombinant WT or R534X NARS1 proteins. We found that the melting temperature (T_m) of the recombinant WT apoenzyme at 48.5°C ± 1.1°C was not significantly different from the R534X apoenzyme, with melting temperatures determined to be 49.3°C ± 0.6°C, respectively (Figure 2A). We also assessed if the addition of substrates, ATP or Asn, altered the T_m for recombinant WT or R534X proteins. The addition of ATP or Asn led to no significant difference in the T_m in homodimeric proteins.

Figure 2.

NARS R534X leads to subtle changes in protein structure and thermal stability

(A) Melting temperature of recombinant NARS1 WT compared to R534X or an equal molar mixture of WT and R534X with or without 1 μM ATP, 1 μM asparagine (Asn), or in combination of 1 μM ATP and 1 μM Asn was determined by a protein thermal stability assay. (B) Limited tryptic digestion of NARS1 WT and NARS1 R534X proteins indicates more rapid degradation of R534X. Error bars represent one standard deviation from at least quadruplicate analysis. C0 and C120 are proteins incubated without trypsin at 0 and 120 min. Significance was calculated by one-way ANOVA, ns, not significant, ∗∗p < 0.01. n = 4.

Since R534X is a dominant allele and has only been identified in heterozygous patients, we next assessed if an equimolar mixture of the WT and R534X proteins causes a shift in the T_m. We found no difference in T_m when comparing the mixture to either WT or R534X alone, but the WT/R534X mixture showed significantly increased T_m upon addition of ATP compared to the heterodimer apoenzyme (Figure 2A). This suggests that the WT/R534X heterodimer adopts a more stable conformation upon binding ATP. R522X was not assessed for thermal stability, as we were unable to purify enough of the recombinant protein due to low levels of protein production, which indicates either a significant change in stability or a protein folding defect of the recombinant protein in E. coli.

Subtle changes in ARS structures can be assessed by limited tryptic digest to probe for differences in surface accessibility, secondary structure, and susceptibility to degradation.^9,28 Proteins were incubated with trypsin, with the reaction quenched at the indicated time points (Figure 2B). The products were separated by SDS-PAGE (Sodium Dodecyl Sulfate Polyacrylaminde Gel Electrophoresis), which shows that the R534X mutant is more rapidly digested compared to the WT protein (Figure 2B) as evidenced by a differential digestion pattern of R534X compared to WT with additional lower molecular weight bands appearing in the R534X digest (Figure 2B, black arrows). These data show that the WT and R534X proteins have distinct tryptic fingerprints and surface accessibility. Overall, our findings indicate that R534X alters protein structure and has a dominant effect on thermal stability in the heterodimer.

Overexpression of dominant NARS1 alleles R522X and R534X causes cytotoxicity

To establish the cellular response of NARS1 nonsense variants in human cells, we generated an overexpression model in HEK293T cells. HEK293T cells were transfected with plasmids expressing either NARS1 WT, NARS1 R522X, or NARS1 R534X. NARS overexpression was achieved at levels comparable to endogenous NARS for the WT allele, as demonstrated by western blotting for NARS. We can clearly distinguish the overexpressed V5-tagged NARS WT from the endogenous NARS, showing almost equal abundance of the overexpressed and endogenous alleles (Figure S1). For the truncated nonsense variants, the V5-tagged R534X and R522X variants run at approximately the same relative molecular weight as the V5 increases the molecular weight to that of endogenous NARS. We then assessed cytotoxicity and cell viability of these cell lines (Figures 3A and 3B). We measured cell viability and cytotoxicity in two separate assays. The cell viability assay (CCK8) measures metabolic activity as a proxy for the number of viable cells, which reflects cell survival. In contrast, the cytotoxicity assay directly measures cell death. No impact on cell viability was observed for overexpression of any of the pathogenic alleles compared to WT NARS1-expressing cells (Figure 3A), but we observed a clear and significant phenotypic defect for the dominant alleles R522X and R534X when measuring cytotoxicity (Figure 3B). For R522X, a significant increase in dead cells over live cells was observed, where the proportion of dead/live cells was increased from 16% to 21%. Similarly, the proportion of dead cells was significantly increased from 16% to 22% in R534X-expressing cells (Figure 3B), indicating toxicity of R534X. When we blotted for R534X expression in the cells using an N-terminal V5 tag, we found that R534X abundance is reduced compared to cells expressing NARS1 WT (Figure 3C), indicating reduced protein stability of R534X in cells.

Figure 3.

Overexpression of NARS1 disease-associated alleles is cytotoxic in HEK293T cells and activates the integrated stress response

(A) CCK8 assay to quantify the number of metabolically active cells and (B) CytoTox-Glo assay to quantify cell death were carried out 24 h after transfection of HEK293T cells with plasmids encoding NARS1 wild-type (WT) or NARS1 R522X and R534X alleles. (C) Expression of WT and R534X NARS1 was confirmed by western blotting against the N-terminal V5 tag. (D and E) Activation of the integrated stress response was determined by (D) western blotting and (E) quantification of blots using ImageJ for abundance of eIF2⍺ and phosph-eiF2⍺ in untreated control cells (CO) and NARS1 WT-, R522X-, and R534X-expressing cells. (F) RT-qPCR of mRNAs normalized to beta-actin that are regulated by the integrated stress response. Error bars represent one standard deviation from at least triplicate experiments. Significance was calculated by one-way ANOVA, where ∗p < 0.05, ∗∗∗p < 0.001, ns, not significant. n ≥ 3.

NARS R522X and R534X expression activates the ISR in HEK293T cells

Previous studies have shown that pathogenic ARS alleles can cause the activation of the ISR.²² The ISR can be activated by the accumulation of uncharged tRNA in the cell, which binds to GCN2, and in turn, phosphorylates eIF2α, initiating transcriptional changes to adapt to nutrient starvation. It has also been shown that ISR activation can arise independently of the interaction of uncharged tRNA with GCN2, where it is mediated through the interaction of the ribosomal P-stalk with GCN2 instead.^29,30 Uncharged tRNA can accumulate either due to amino acid starvation or due to ineffective ARS aminoacylation, even in the presence of sufficient amino acid substrates.^{21,23,24,28,31} We, therefore, probed for eIF2α abundance and phosphorylation. We indeed found that relative phosphorylation of eIF2α was significantly increased in R522X and R534X compared to WT NARS1-expressing cells (Figures 3D and 3E). The abundance of eIF2α was not changed (Figure S2A), but significantly increased phosphorylation was evident for the nonsense NARS1 mutants (Figure S2B). We also assessed changes in the expression of transcription factors known to be regulated by the ISR. We found no significant changes in transducin beta-like 2 (TBL2), activating transcription factor 3 (ATF3), or ATF4 RNA abundance (Figures 3F and S1), which may indicate that ISR signaling is not fully activated at 24 h. In addition, we probed for RNA abundance of neuropilin 1 (NRP1), a receptor which is involved in nervous system development and shown to interact with CMT-causing variants of AARS1 and GARS1,^32,33 and found no changes in NRP1 mRNA abundance (Figure 3F). Overall, these data indicated that dominant NARS1 nonsense alleles elicit cytotoxicity and activate the ISR in human cells.

Overexpression of NARS1 R522X perturbs the proteome

To assess if overexpression of pathogenic NARS1 alleles alters the composition of the proteome, we carried out a proteomic analysis of cells overexpressing either WT NARS1 or NARS1 R522X. We found that several proteins were differentially abundant in R522X-overexpressing cells (Figure 4A), and a principal-component analysis showed that the proteomes of NARS1 and NARS1 R522X-overexpressing cells are distinct (Figure 4B). When using a 2-fold cutoff, we found that 4 proteins were significantly reduced in R522X compared to WT cells (Figure 4A). When decreasing the cutoff to 1.8-fold, we found that 49 proteins are not detected or significantly reduced in R522X-expressing cells compared to WT, and 57 proteins are increased in abundance (Figures 4C and 4D; Table S2, Supplementary File 1). A STRING and Gene Ontology (GO)-term analysis of the up- or downregulated proteins separately did not identify enrichments in specific pathways (Figure S3). Combining both up- and downregulated proteins into a STRING search identified mRNA processing/splicing (KW:0507, mRNA splicing, count in network 8/349, false discovery rate [FDR] 0.0137) and RNA binding proteins (GO:0003723, RNA binding, count in network 24/1627, FDR 0.0113) as significantly enriched pathways and cellular functions (Figure 4D). Notably, several proteins associated with the 60S ribosomal subunit are dysregulated. Ribosome biogenesis protein BRX1 homolog (BRX1) and ribosomal protein L3-like (RPL3L) are not detected in the WT NARS1-expressing samples but are detected abundantly in all three R522X-expressing cell samples (Table S2). Conversely, several other ribosome-associated proteins are downregulated in R522X cells. A subunit of the 60S ribosome, ribosomal protein L24 (RPL24), was downregulated 4-fold, and the regulator of ribosome synthesis 1 (RRS1) was downregulated by 3-fold in R522X cells. Interestingly, the transcription factor specificity protein 1 (SP1), which is involved in regulating cell differentiation, growth, apoptosis, response to DNA damage, and chromatin remodeling, is also undetectable in WT cells but abundant in two of the three R522X replicates (Table S2). The proteomic data indicate that in response to NARS1 R522X, RNA processing and protein synthesis pathways are most impacted, showing a dysregulation of cellular and protein homeostasis. No mistranslation above background was identified in the R522X cells, suggesting that translation accuracy is not affected by NARS1 R522X.

Figure 4.

Overexpression of NARS1 R522X causes a perturbation of the proteome

(A) Volcano plot and (B) principal-component analysis of proteins significantly up- (red) or downregulated (green) in NARS1 R522X-expressing cells compared to WT NARS1-expressing cells by at least 2-fold. (C) Heatmap of the (D) protein-protein interaction network of proteins differentially regulated by at least 1.8-fold generated with STRING.³⁴n = 3.

NARS1 cytotoxicity is rescued by substrate supplementation

We previously showed that supplementation with cognate amino acid and tRNA substrate can rescue pathogenic phenotypes in a humanized Saccharomyces cerevisiae model of CMT disease arising from mutations in the histidyl-tRNA synthetase (HARS1).^9,11,12 We, therefore, investigated the ability of Asn or tRNA^Asn supplementation to rescue the cytotoxic phenotype of NARS1 R534X and NARS1 R522X alleles.

At 24 h post-transfection, R522X and R534X elicit a cytotoxic phenotype compared to WT (Figure 5A). In untreated control cells, Asn supplementation did not affect cell viability or cytotoxicity (Figures 5A and 5B). Surprisingly, tRNA supplementation alone had a cytotoxic effect in control experiments, increasing the proportion of dead cells from 9.5% to 14% in cells treated with tRNA^Asn compared to transfection reagent only controls (Figure S4). Since the expression levels of tRNA^Asn were not tested, there may be a dose-dependent effect that may affect toxicity and can be tested in future experiments. We next tested whether Asn or tRNA^Asn supplementation is cytotoxic to WT NARS1-expressing cells, but no change in cell viability or cytotoxicity was observed (Figures 5A and 5B). Excitingly, when supplemented with 200 mg/L Asn, we found reduced cytotoxicity for each of the nonsense variants, with R522X and R534X showing pronounced rescue after 24 h to levels indistinguishable from WT NARS-expressing cells (Figure 5A). Similarly, when co-expressing tRNA^Asn from the same plasmid as the NARS1 allele, we also found cytotoxicity to be significantly suppressed in cells expressing the R522X and R534X alleles. These data demonstrate that substrate supplementation with either the cognate tRNA or the cognate amino acid can rescue the cytotoxic phenotype of the NARS1 nonsense variants.

Figure 5.

Rescue of NARS1 nonsense allele cytotoxicity with asparagine, human tRNA^Asn, or an inhibitor of the integrated stress response

(A) CytoTox-Glo assay to quantify cell death and (B) CCK8 assay to quantify the number of metabolically active cells were carried out 24 h after transfection of HEK293T cells with plasmids encoding NARS1 wild-type (WT), NARS1 R522X, or NARS1 534X. Cells were incubated with and without 1 nM of the GCN2 inhibitor ALE2 and 200 mg/L asparagine (Asn) or supplemented with tRNA^Asn expressed from the same plasmid as the NARS1 gene. Activation of the integrated stress response was determined for R534X (C and D) and R522X (E and F) by (C and E) western blotting and (D and F) quantification of blots using ImageJ for abundance of eIF2⍺ and phospho-eiF2⍺ in cells and calculated as relative abundance of phospho-eIF2⍺. The impact of treatments on transgene expression was assessed via western blotting (G and H) and quantification (I and J) of V5-tagged WT (G and I) and R534X (H and J). Error bars represent one standard deviation from at least triplicate experiments. Significance was calculated by one-way ANOVA, where ∗p < 0.05, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns, not significant. n ≥ 3.

Asn supplementation ameliorates activation of the ISR

Activation of the ISR itself is associated with CMT and other disease phenotypes.^22,24,35 To determine whether substrate supplementation impacts stress response signaling, levels of total eIF2α and phosphorylated eIF2α (p-eIF2α) were assessed by western blotting. As mentioned earlier, expression of the R534X mutant led to elevated levels of p-eIF2α, consistent with activation of the ISR (Figures 5C and 5E). Western blots were also performed following supplementation of the transfected cells with Asn or co-expression of tRNA^Asn to assess whether these would reduce eIF2α or p-eIF2α abundance and relative phosphorylation levels. We also treated the control cells with halofuginone to demonstrate a full activation of the ISR,³⁶ which led to a 2-fold increase in p-eIF2α (Figures 5C–5F). Under all conditions, no change in eIF2α abundance was observed relative to a tubulin loading control (Figures 5C–5F; Figures S1C and S1E). R534X cells show a 2-fold increase in absolute and relative p-eIF2α phosphorylation, and we observed a reversal of eIF2α phosphorylation upon Asn supplementation (Figures 5D and S2D) to levels indistinguishable from cells expressing WT NARS1. The data show that Asn supplementation effectively suppresses NARS1 R534X-induced ISR stimulation.

For the R522X variant, the effect on the activation of the ISR is less clear. While the ISR is significantly activated in R522X cells, we found that the therapeutic rescue approaches with tRNA^Asn and Asn led to eIF2α phosphorylation levels that were not significantly different from either the WT or the R522X cells (Figures 5E, 5F, and S2F). Together, these results demonstrate that while NARS1 nonsense variant cytotoxicity can be rescued by providing exogenous Asn or tRNA^Asn, the ISR stimulation is only rescued by Asn supplementation in both nonsense variants.

NARS1 cytotoxicity is rescued by inhibition of the ISR

Previous work, as well as our results here, demonstrates that the ISR is activated in ARS-related disease models.²² ISR inhibitors, including those that target GCN2, would serve as a potent treatment option if proven efficacious.^37,38 We tested a newly developed GCN2 inhibitor provided by Alesta Therapeutics (ALE2) in our NARS1 cell models. We found that the NARS1 R522X and R534X-induced cytotoxicity was dramatically reduced after 24 h following ALE2 treatment, with the proportion of dead cells reduced from 21% to 10% for R522X and 22% to 13% for R534X (Figure 5A).

Since GCN2 phosphorylates eIF2α, we probed eIF2α abundance and phosphorylation by western blotting but found no significant reduction in p-eIF2α levels (Figures 5C and 5D). Similar to the tRNA^Asn and Asn supplementation for the R522X model, a reduction in p-eIF2α levels was observed upon ALE2 treatment (Figures 5E and 5F), which reduced p-eIF2α phosphorylation to a level indistinguishable from both WT and R522X. Lastly, to assess whether our proposed therapeutic interventions impact the transfection efficiency, and subsequent recombinant NARS expression in our model, western blotting was performed, probing for V5-tagged NARS post-treatment. No changes in transgenic NARS abundance were observed across treated and untreated conditions for the WT (Figures 5G–5I) and R534X variant (Figures 5H and 5J), with the exception of WT NARS co-expression with tRNA^Asn, which resulted in a 40% reduction in V5-NARS abundance.

Impact of therapeutic approaches on the human proteome

Since we observed a perturbation of the proteome for cells expressing the NARS1 R522X variant, we also carried out a proteomic analysis of the phenotypic rescues with Asn, tRNA^Asn, and the GCN2 inhibitor ALE2 (Figures 6A–6C; Figure S3). We used PEAKS software to identify proteins that are significantly changed in at least one condition by at least 2-fold and analyzed these proteins using STRING (Figures 6D and S3). We identified several pathways, including translation, ribosome biogenesis, vascular endothelial growth factor (VEGFA/VEGFR) signaling, RNA metabolism, and RNA binding proteins, as most impacted (Figure 6D; Table S2). When comparing the R522X proteome to both the WT and the different rescue approaches, we found that of the three treatments, Asn supplementation led to the most restored proteome. Using a 2-fold cutoff, there are 41 differentially abundant proteins when comparing WT to R522X cells, and this number is reduced to 32 differentially abundant proteins when comparing WT to Asn-supplemented R522X cells (Figures 6E and S3; Table S2). In contrast, when comparing the tRNA-supplemented cells, we found additional proteins upregulated in the tRNA-supplemented cells. tRNA supplementation of R522X cells increased the number of differentially abundant proteins to 69 proteins compared to WT (Figures 6E and S3; Table S2), with 49 proteins differentially regulated between R522X-supplemented and unsupplemented cells (Figures 6E and S3H). Finally, GCN2 treatment of R522X cells led to differential regulation of 64 proteins compared to WT NARS-expressing cells and 84 proteins compared to R522X cells (Figures 6E and S3H; Table S2). These data show very clearly that the proteome of Asn-supplemented R522X cells most closely resembles the WT NARS proteome, with the least changes between WT and Asn-supplemented R522X cells (Figure 6E).

Figure 6.

Therapeutic approaches to restore the NARS1 R522X proteome

Volcano plots of WT NARS1-expressing cells compared to (A) R522X cells supplemented with asparagine (Asn), (B) tRNA^Asn, or (C) the GCN2 inhibitor ALE2. (D) Protein-protein interaction network of proteins significantly different in abundance in at least one condition by at least 2-fold generated using STRING.³⁴ (E) Proteins significantly up- or downregulated by at least 2-fold when comparing indicated proteomes. n = 3.

We found several pathways impacted by the R522X allele. For example, ribosome biogenesis is impacted in R522X-expressing cells (Table S2), as exemplified by the downregulated RRS1 and the upregulated ribosomal protein S17 (RPS17). Interestingly, RRS1 levels are restored to WT levels using all three therapeutic approaches (Figure 7A). RPS17, a protein component of the 4Ss subunit, is upregulated in R522X cells and, interestingly, is further upregulated in GCN2 inhibitor-treated or Asn-supplemented cells, while tRNA supplementation reduced RPS17 levels below WT levels (Figure 7B). Peptidyl-tRNA hydrolase 2 (PTRH2) releases tRNAs from the ribosome during protein synthesis and is upregulated by 2.5-fold in R522X cells, with levels restored to WT levels in Asn or GCN2 inhibitor-treated cells (Figure 7C). These data support the expected consequence of globally impaired translation due to activation of the ISR and is further reinforced by the enrichment and upregulation of proteins involved in the negative regulation of gene expression. This includes proteins such as cold shock domain-containing E1 (CSDE1), a protein involved in the RNA-induced silencing complex (RISC) formation,³⁹ which was found to increase nearly 2-fold in R522X-expressing cells compared to WT (Table S2). Overall, these data show a significant impact on translation components in R522X cells, which can mostly be restored by Asn supplementation.

Figure 7.

Asparagine supplementation rescues changes in the R522X proteome

Peptide area obtained from the proteomic analysis for (A) regulator of ribosome synthesis 1 (RRS1), (B) ribosomal protein S17 (RPS17), (C) peptidyl-tRNA hydrolase 2 (PTRH2), (D) F-actin-capping protein subunit beta (CAPZB), (E) catenin delta-1 (CTNND1), (F) dead-box protein 3 (DDX3X), (G) X-ray repair cross complementing 5 (XRCC5), and (H) angiomotin (AMOT). Significance was calculated by one-way ANOVA, where ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns, not significant. n ≥ 3.

In addition to translation-associated proteins, we also identified proteins with a function in cell structure and adhesion to be deregulated in cells expressing mutant NARS1. Catenin delta-1 (CTNND1) encodes the p120-catenin (p120ctn) protein with a function in cell adhesion,⁴⁰ while F-actin-capping protein subunit beta (CAPZB) is a member of the F-actin capping protein family and regulates actin filament growth.^41,42 Both proteins are upregulated in R522X cells, and abundance is restored to WT levels in Asn-supplemented cells (Figures 7D and 7E). Interestingly, tRNA supplementation further increases CTNND1, while GCN2 inhibition reduces CAPZB.

We found several proteins that are associated with neurodevelopmental disorders, including the VEGF signaling pathway, which is altered in response to R522X expression. The VEGF pathway regulates both blood vessel formation (angiogenesis) and neuron development (neurogenesis),^43,44 and its disruption is known to contribute to developmental disorders.^44,45 In addition, the VEGF pathway has shown to be indirectly linked to ISR activation, specifically through increased expression of the stimulatory VEGF ligand in response to endoplasmic reticulum (ER) stress and amino acid deprivation, which are conditions synonymous with the ISR.⁴⁶ Examples of proteins in the VEGF pathway include dead-box protein 3 (DDX3X) and X-ray repair cross complementing 5 (XRCC5). DDX3X is an RNA helicase involved in splicing and associated with DDX3X-related neurodevelopmental disorder, which presents with similar symptoms to NARS1 disease, including microcephaly, movement disorders, and seizures.⁴⁷ XRCC5 encodes the Ku80 protein, an integral part of the Ku70/80 heterodimer that is part of non-homologous end joining and DNA repair and is indirectly involved in VEGF signaling.⁴⁸ Both XRCC5 and DDX3X are downregulated in R522X cells, but abundance is restored in all three therapeutic approaches (Figures 7F and 7G). Angiomotin (AMOT) is a crucial suppressor of cancer metastasis and interacts with the VEGF signaling pathways.⁴⁹ AMOT is upregulated by 2-fold in R522X cells and returned to WT levels in Asn-supplemented cells. Interestingly, tRNA supplementation of R522X-expressing cells further increases AMOT abundance to 8-fold compared with WT. GCN2 inhibition treatment of NARS R522X cells, on the other hand, reduces AMOT to levels below WT (Figure 7H). Together, these data show that VEGF signaling is perturbed in the R522X cells and that the proteomic changes can consistently be reversed with Asn supplementation, while tRNA supplementation and GCN2 inhibition only restore some of the changed protein levels.

We also checked for changes in heat-shock protein abundance and ARS levels, as changes in these have been observed in other model systems for ARS diseases,^2,9,11,12 but found no significant changes related to these pathways in the detected proteins between WT- and R522X-expressing cells. Interestingly, we found increased abundance of two ARS proteins, valinyl-tRNA synthetase and bifunctional glutaminyl-prolyl tRNA synthetase, in tRNA-supplemented cells, indicating a proteomic response to the perturbation of the tRNA pool (Table S2; Figure S5).

Asn supplementation rescues a viability defect in R534X patient-derived fibroblasts

While our overexpression model in HEK293T cells revealed a defined phenotypic defect, we next sought to confirm the efficacy of Asn supplementation in patient-derived fibroblasts derived from a patient with a heterozygous R534X allele. Similar to other case reports of individuals with the R534X variant,² the individual who donated the fibroblast cells displayed symptoms of developmental delay, microcephaly, intellectual disability with coordination issues and delayed speech development, an abnormal electroencephalogram (EEG) with seizures, and mixed hypotonia and hypertonia within the first year of life. At the time of the fibroblast sample acquisition, the patient’s seizures had resolved; however, peripheral neuropathy was diagnosed. Additional symptoms in this individual not previously documented in the literature include ongoing issues with eating-related gagging and coughing, as well as autism with stereotypes, hyperactivity, and problems with appropriate communication, along with sleep disruption with nighttime awakenings.

Interestingly, we found that the patient-derived fibroblasts did not show a cytotoxic phenotype but rather a significant defect in cell viability compared to the control fibroblasts. R534X fibroblasts display significantly reduced viability, showing 18% reduced overall viability compared to the control cells (Figure 8A). In contrast, R534X fibroblasts had significantly reduced cytotoxicity compared to control fibroblasts, which may be attributed to slower overall cell growth (Figure 8B). Next, we supplemented the media with Asn or ALE2 and a DMSO-only control for the ALE2-treated cells. To assess tRNA supplementation, we transfected the fibroblasts with a tRNA^Asn overexpression plasmid and used a transfection reagent-only control. We also tested the supplementation of cells with glucose, with the intent of increasing cellular ATP production from glucose to overcome a potential defect in the first step of aminoacylation.⁵⁰ Glucose and ALE2 were found to have no impact on the viability of R534X fibroblasts (Figure 8A). tRNA^Asn supplementation was achieved by transfection of a plasmid encoding the tRNA, compared to cells treated with transfection reagent only. Transfection reagent only ameliorated the difference in patient from control cells, with no difference observed between the cell lines, potentially due to the cytotoxic effect of lipofectamine that reduced overall cell growth in both cell lines.⁵¹ Nonetheless, tRNA supplementation was ineffective in the fibroblast model as it significantly reduced viability for both control and patient-derived cell lines (Figure 8A). Most notably, when supplemented with Asn, viability and cytotoxicity in R534X fibroblasts were indistinguishable from the control fibroblasts (Figures 8A and 8B).

Figure 8.

Asparagine supplementation rescues a cell viability defect in NARS1 R534X patient-derived fibroblasts

(A) CCK8 assay to quantify the number of metabolically active cells and (B) CytoTox-Glo assay to quantify cell death were carried out 24 h after plating patient-derived NARS1 R534X fibroblasts and age-matched control fibroblasts 24 h following seeding at an initial density of 25,000 cells. Cells were untreated (control, CO), supplemented with asparagine (Asn) or glucose, or transfected with an empty plasmid (TR) or a tRNA^Asn-expressing plasmid (tRNA). (C and D) Aminoacylation assay in cell extracts from patient fibroblasts or control fibroblasts under (C) high Asn or (D) stress conditions with increased temperature and reduced Asn availability. (E) Western blots and (F) quantification of proteins involved in the integrated stress response in CO and Asn-supplemented patient fibroblasts or control fibroblasts. (G) RT-qPCR of mRNA transcripts normalized to beta-actin that are controlled by the integrated stress response, as well as NARS1. (H–K) Asn is not readily turned over into aspartate, glutamate, or glutamine and excreted when supplemented in an HEK293T overexpression model. Relative quantification of (H) Asn, (I) glutamine, (J) glutamate, and (K) aspartate was performed using 1D ¹H-NMR experiment from spent media of HEK293T cells expressing R522X or WT NARS1 with and without supplementation of Asn for 24 h. The y axis represents relative peak intensity from 1D ¹H-NMR spectra of the indicated metabolites. Histogram values were normalized to unsupplemented WT NARS-expressing cells. Error bars represent one standard deviation from at least triplicate experiments. Significance was calculated by one-way ANOVA (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns, not significant). n ≥ 3.

R534X was previously suggested to lead to reduced aminoacylation activity.^2,17 We therefore measured NARS1 activity in the R534X patient-derived cells, comparing aminoacylation under high and low (stressed) conditions.⁵² Asn concentrations for the stressed condition were set at 0.075 mM. Reference plasma values in patients are usually between 0.03 and 0.085 mM. The experimental Asn concentration was set at 0.3 mM Asn, and even under these high Asn conditions, NARS1 activity was reduced to 25% in the patient-derived cells compared to control fibroblasts, indicating starkly reduced aminoacylation in NARS R534X cells. Under stress conditions, using increased temperature and reduced Asn availability, the disparity became even more evident, reducing NARS1 R34X aminoacylation activity to 8% of the control fibroblasts (Figures 8C and 8D). These data highlight that Asn availability is crucial for efficient aminoacylation in patient-derived cells. These findings suggest that Asn supplementation improves cell viability in the patient-derived fibroblasts and increases growth and turnover to a rate comparable with unaffected cells, suggesting Asn supplementation as a potential therapeutic approach for these patients.

Asn supplementation prevents transcriptional activation of ATF3 and ATF4 in patient-derived fibroblasts

To assess whether the ISR is activated in patient-derived fibroblasts as it was in the HEK293T overexpression model, western blotting and RT-qPCR were performed on fibroblast lysates and RNA extracts. Untreated control fibroblast mRNA extraction yielded slightly more mRNA, and we normalized all values to beta-actin (Figure S6). In contrast to the HEK293T overexpression model, p-eIF2α and eIF2α levels were not suggestive of activation of the ISR, with quantification showing that the abundance of both eIF2α proteoforms in the R534X fibroblasts was not significantly different compared to the control fibroblasts (Figures 8E and 8F). Nonetheless, quantification of the downstream transcription factors that are regulated during the ISR showed that ATF3 and ATF4 transcripts were significantly increased by 2.3- and 2.6-fold compared to control cells (Figure 8G), indicating an overall upregulation of the ISR. Upon supplementation with Asn, ATF3 and ATF4 levels returned to levels indistinguishable from control cells, indicating a rescue of the ISR activation (Figure 8G).

Transcripts with premature stop codons are often subject to nonsense-mediated decay (NMD).^53,54 We therefore quantified NARS1 mRNA abundance by RT-qPCR and surprisingly found a 2.2-fold increase in NARS1 transcript abundance (Figure 8G), suggesting that the NARS1 R534X mRNA is not a substrate for NMD. Quantitative PCR cannot distinguish between the WT allele and the R534X allele, and it is possible that transcription of the WT NARS1 allele is upregulated, perhaps in a compensatory attempt to stabilize NARS1 protein levels. Interestingly, the NARS transcript abundance returned to WT levels upon Asn supplementation, indicating that increased NARS expression is a compensatory effect in the patient-derived cells. Since no increase in NARS1 protein abundance was found in previous studies for R534X fibroblasts relative to control cells,² these data indicate that the increase in mRNA abundance may indeed compensate for reduced R534X protein abundance we observed in R534X-overexpressing cells (Figure 2C).

One of the concerns with using Asn as a treatment for a neuropathological disease, such as NARS1 disease, is its metabolization into aspartate and conversion in the Asn synthetase (ASNS) pathway into glutamate. Glutamate is an excitatory neurotransmitter that can worsen seizures and patient outcomes when found in excess or when supplemented orally.^55,56,57 We, therefore, employed nuclear magnetic resonance (NMR) spectroscopy to determine if Asn is readily converted to glutamate or other metabolites in the ASNS pathway and secreted to the media to act as a potential neurotransmitter. Spent media from HEK293T cells overexpressing WT or R522X NARS1 were collected 24 h post-transfection with or without treatment with 200 mg/L Asn and analyzed by NMR. As expected, the Asn peak intensity shows a significant increase from the spectra generated from the treated WT and R522X samples compared to the untreated samples (Figure 8H). No significant difference in peak intensity was observed in the spectra of glutamate, glutamine, or aspartate, across all treated and untreated groups (Figures 8J and 8K), indicating no increased turnover of Asn into aspartate, glutamate, or glutamine. Amino acid levels should nonetheless be monitored if amino acid supplementation is considered in patients. Overall, these findings reveal that the R534X mutation results in decreased aminoacylation activity and cellular viability in patient-derived fibroblasts. These NARS1 phenotypic defects can be rescued by Asn but not by cognate tRNA supplementation or GCN2 inhibition.

Discussion

The consequences of pathogenic NARS1 alleles on patients are severe and have recently been compiled into a comprehensive disease description.¹ While progress has been made in identifying the genetic cause, the underlying molecular mechanism for most of these alleles remains unclear, and there is currently no disease-specific treatment available. We here investigated the molecular and cellular mechanisms of two dominant NARS1 alleles R522X and R534X and tested potential therapeutic treatments, using cell-based and biochemical assays.

Molecular and cellular consequences of pathogenic NARS1 alleles

Previous in vitro studies have also shown that the recombinant R534X NARS1 protein exhibits reduced aminoacylation activity and leads to dysfunctional heterodimer formation and that substrate binding may destabilize dimer formation as we also observed here.¹⁷ Our data show that both nonsense variants R522X and R534X lead to a cytotoxic phenotype in HEK293T cells, and we found reduced cell viability in patient-derived fibroblasts. It should be noted that the observed phenotypes between HEK293T cells and patient-derived fibroblasts were not always consistent, where e.g., in contrast to fibroblasts, no defect in cell viability is observed in HEK293T cells, but we instead observed cytotoxicity. This discrepancy can likely be attributed to the differences in genetic backgrounds between HEK293T cells and fibroblasts lines, as well as diverse genetic backgrounds of patient and control lines. In addition, HEK293T cells retain multiple (up to three due to chromosomal instability) WT NARS alleles. Nonetheless, we found that the dominant-negative nonsense alleles showed a phenotype in both models. In addition, a previous study showed that the R534X allele demonstrates a dominant-negative growth phenotype in a yeast model, which aligns with the dominant phenotype we observed here.¹⁶ Our data from patient-derived fibroblasts confirm that heterozygous NARS1 R534X fibroblasts retain less than 25% NARS1 aminoacylation activity, which is further reduced under stress conditions. Overexpression of NARS1 R534X in human cells shows reduced protein abundance compared to the WT NARS1, indicating reduced stability of NARS1 R534X in the context of human cells. Interestingly, we also observed reduced protein abundance for WT NARS1 when co-expressed with tRNA^Asn. This may be attributed to NARS being actively degraded to prevent the production of excessive Asn-tRNA^Asn, although this remains speculative. We previously showed that some pathogenic variants in HARS1 cause mistranslation,^9,12 but we found no indication of mistranslation generated by mutant NARS1 in our proteomic experiments. This, as well as previously published data,^2,4,17 indicated that a dominant loss of function may be the cause of the observed phenotypic and cytotoxic effects of the dominant NARS1 alleles.

Our proteomic data show several changes in the proteome of NARS1 R522X-overexpressing cells. Some of these changes, and the lack thereof when compared to our treatment conditions, may be attributed to variation between replicates. Western blotting probing for specific proteins associated with our observed proteomic perturbations may be performed in the future to further validate these findings. It is interesting to note that a more dramatic effect was observed when comparing the transcriptome of patient-derived fibroblasts with bi-allelic NARS variants to control fibroblasts.² These sequencing data showed increased expression of several other ARSs in homozygous and compound heterozygous patient-derived fibroblasts, including cytoplasmic glutamyl-tRNA synthetase (DARS1), GARS1, cytoplasmic arginyl-tRNA synthetase (RARS1), cytoplasmic seryl-tRNA synthetase (SARS1), cytoplasmic threonyl-tRNA synthetase (TARS1), cytoplasmic tryptophanyl-tRNA synthetase (WARS1), and cytoplasmic tyrosyl-tRNA synthetase (YARS1).² In contrast, we observed no significant changes in ARS proteins in the proteome of the dominant NARS R522X allele. Despite larger transcriptomic changes in the bi-allelic variants, we found that the deregulated pathways largely overlap with our proteomic observations. Like our data, these transcriptomic analyses of bi-allelic NARS1 variants revealed upregulation of VEGF-associated pathways, upregulation of the ISR (e.g., ATF4), and changes in ribosomal homoeostasis.² The ISR is also activated in other synthetase-associated diseases, such as GARS disease models in mice^21,22,38 and HARS disease models in rat cells.⁵⁸ These data show that while each ARS and even each specific ARS allele can cause a variety of different phenotypes and diseases, some of the cellular responses may reveal common themes, such as ISR activation.

Therapeutic approaches to NARS1 disease

Mechanistically, it is thought that supplementation with cognate amino acid or cognate tRNA will increase enzyme activity of an aminoacylation-deficient ARS to a level that is sufficient to sustain translation or outcompete non-specific binding events that lead to mistranslation. Reduced ARS activity can lead to uncharged tRNA accumulation, which activates the ISR.⁵⁹ Uncharged tRNA can directly stimulate ISR activation through binding to GCN2 via its HARS-like domain, or, as discovered more recently, can act indirectly by allowing GCN2 to interact with the ribosomal P-stalk, which also serves as a potent trigger of this stress response.^29,30 Restoring sufficient and correct aminoacylation activity would prevent the activation of the ISR, which can itself be pathogenic.^21,22,23,35

Over the past few years, amino acid supplementation has gained increasing attention in the treatment of ARS-associated diseases.^60,61,62 For example, methionine supplementation led to phenotypic improvement in an infant and children affected with cytoplasmic methionyl-tRNA synthetase (MARS1)-associated diseases,^62,63,64 allowing for a successful lung transplantation in three methionine-supplemented patients with MARS-related pulmonary alveolar proteinosis.⁶⁵ In another success story, patients with bi-allelic variants in Isoleucyl-tRNA synthetase (IARS), leucyl-tRNA synthetase (LARS), phenylalanyl-tRNA synthetase subunit B (FARSB) and seryl-tRNA synthetase (SARS) were treated with the corresponding amino acids, resulting in overall improvement and management of symptoms.⁶⁶ A mitochondrial phenylalanyl-tRNA synthetase (FARS2) deficiency responded to phenylalanine supplementation.⁶⁷ These and other examples indicate amino acid supplementation as a first-line treatment for many ARS-associated diseases.^59,66 Indeed, we found that supplementation of Asn rescued not only cytotoxicity but also reversed many of the changes in the proteome of NARS1 R522X-overexpressing cells and rescued cell viability and aminoacylation defects in patient-derived fibroblasts. On the other hand, while ATP stabilized the recombinant protein, increasing glucose concentrations in the media with the intent to increase ATP availability in the cell did not lead to a phenotypic rescue. Together, these findings indicate that Asn should be further studied and clinically evaluated as a potential first-line treatment for NARS1 patients.

While we are hopeful that Asn supplementation will reverse symptoms in NARS1 patients or at least slow progression, as previously observed for other ARS-associated diseases noted earlier, it is also important to consider the development of additional therapeutics. Our previous work in humanized yeast models showed that amino acid supplementation may not be a one-size-fits-all therapeutic approach. For example, histidine was shown to rescue some HARS alleles, while other alleles, even near the active site, can be exacerbated by histidine supplementation.¹² These variants were rescued by cognate tRNA supplementation in yeast.⁹ The cognate tRNA supplementation approach was also successful in restoring a mouse phenotype modeling GARS1-associated disease.⁶⁸ Indeed, tRNA therapeutics have gained interest, and the development of therapeutic tRNA delivery strategies has brought tRNA medicine closer to clinical applications.^59,69 Our data show that cognate tRNA supplementation indeed rescued the cytotoxic effect of pathogenic NARS1 variants but did not ameliorate ISR activation, though the interpretation of the result is complicated by the fact that the level of tRNA^Asn overexpression is not known. Perhaps the finding is not surprising, since delivering additional uncharged tRNA facilitated a rescue of NARS1 activity to levels sustaining growth by providing more Asn-tRNA substrate for translation; this approach may also provide additional uncharged tRNA to trigger GCN2 activation. While tRNA supplementation may be less desirable for variants that respond to Asn supplementation, future experiments will show if tRNA supplementation may be a therapeutic approach for other NARS1 alleles. For example, an in silico analysis of the R545C mutant NARS1 protein using a nematode ortholog suggests impaired tRNA binding,² which may be rescued by tRNA rather than amino acid supplementation. Another interesting approach for the rescue of nonsense alleles such as NARS R522X and R534X, but also many other nonsense ARS alleles, may be found in tRNA nonsense suppressors, allowing for the production of the full-length NARS protein. This approach was shown to be successful for other disease-causing nonsense alleles, and we will explore nonsense suppression in the future.^69,70,71

tRNA and amino acid supplementation are highly promising approaches for the treatment of ARS-associated diseases but nonetheless may not be successful for all alleles or require further supporting pharmaceutical intervention. The chronic activation of the ISR has been identified as contributing to ARS-associated and other diseases.^22,23,35 Inhibitors of GCN2 have been considered therapeutic options for several diseases, including cancer.^{36,38,72,73,74} The reduction of ISR markers is a hallmark measurement for the rescue of multiple ARS-related diseases.^{5,19,21,22,28,31} Previous work in mouse and yeast models of ARS-related diseases has demonstrated that treatment with GCN2 inhibitors, cognate amino acid, or tRNA is efficacious in improving pathogenic phenotypes.^9,11,12,22 Our data demonstrate that the GCN2 inhibitor ALE2 can rescue a cytotoxic phenotype caused by pathogenic NARS1 alleles in a HEK293T overexpression model. As expected, the inhibitor was unable to restore the dysregulated proteome caused by NARS1 nonsense variant expression. Nonetheless, considering the chronic activation of the ISR is likely a large contributing factor to NARS1 disease development and progression and has previously been associated with the pathogenicity of CMT-causative ARS alleles,^22,23 ISR inhibitors and their continued development represent an important avenue for a small-molecule therapeutic approach for ARS disease.

Conclusion

Genetic diseases, including the extremely rare NARS1 disease, place a large burden on patients and patient caregivers, while drug development is of limited commercial interest to the pharmaceutical industry.⁷⁵ We investigated the cellular consequences of two nonsense NARS1 alleles and found that the supplementation of Asn leads to a rescue of the cellular phenotypes in both human cell models as well as NARS1 patient-derived fibroblasts. Asn supplementation restored the cellular proteome to closely resemble the WT proteome and prevented activation of the ISR. tRNA supplementation and small-molecule inhibition of the ISR also restored the cellular phenotype but only partially rescued ISR activation and the proteome. We previously found that amino acid supplementation may not be a one-size-fits-all treatment for ARS-related disorders and can indeed exacerbate phenotypes for certain HARS mutants. Indeed, different dominant NARS alleles were shown to complement a yeast NARS deletion strain at various degrees, which indicates that not all disease-causing NARS alleles elicit the same cellular consequences.¹⁶ Therefore, personalized medicine approaches need to take allele-specific phenotypes and cellular consequences into account.^9,12 Considering that therapeutic tRNA delivery to patients is still in development and GCN2 inhibitors have not passed safety tests yet, the low-hanging fruit of ARS disorders is amino acid supplementation. Nonetheless, we and others believe that this may only be successful in some cases and may not ameliorate all symptoms in patients. It is therefore valuable to assess alternative strategies, such as cognate tRNA supplementation, nonsense suppression for nonsense alleles, inhibition of the ISR, or other supportive treatments that we have not considered here. For the NARS1 alleles (R534X and R522X) characterized in this study, our data suggest that the readily available supplement Asn may point toward a viable therapy.

Materials and methods

Plasmid preparation and synthesis

The open reading frame for human NARS1 was inserted in the gateway-compatible entry vector pDONR223 (Horizon Discovery, Cambridge, UK). R522X and R534X point mutations were introduced by QuikChange site-directed mutagenesis as described previously.⁷⁶ The NARS1 gene was then cloned into pcDNA3.2-ccdB-V5 (Cat. # 12489019, Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA) for expression in mammalian cells with an N-terminal fused V5 tag, following the standard procedures of Gateway cloning.⁷⁷ For expression of tRNA^Asn, the human tRNA^Asn gene and 300 bases of the flanking up and downstream regions were amplified by polymerase chain reaction (PCR) from HEK293T genomic DNA with primers (Table S1) including terminal PciI (Cat. #R0655S, New England Biolabs, Ipswich, MA, USA) restriction sites. The tRNA cassette was then cloned into pcDNA3.2-ccdB-V5-NARS1 at the PciI site. For recombinant NARS1 expression in E. coli, human NARS1 was cloned into pDEST527 with a His₆ tag for purification using Ni²⁺-nitrilotriacetic acid (Ni-NTA) affinity column chromatography as previously described.^12,18 The sequence of each construct was verified by DNA sequencing at Azenta Life Sciences (Chelmsford, MA, USA). All primers are listed in Table S1.

Mammalian cell culture and assays

HEK293T cells (ATCC #CRL-3216) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (#11995073, Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS) (#A5256701, Gibco, Thermo Fisher Scientific) and 1% penicillin/streptomycin (#15640055, Thermo Fisher Scientific) in a humidified incubator at 37°C with 5% CO₂. For the cell viability assay (CCK8, Abcam, Cambridge, UK) and the cytotoxicity assay (CytoTox-Glo, Promega Corporation, Madison, WI, USA), cells were seeded at a density of 10,000 cells per well in 96-well plates. For RT-qPCR, cells were seeded at 300,000 cells per well of a 6-well dish. Cells were incubated overnight to allow for proper attachment after seeding. For NMR analysis, cells were grown to ∼70% confluency in T75 flasks.

Fibroblasts from either a control individual or a patient with a NARS1 R534X mutation were obtained from skin-punch biopsies and kindly provided by The Rory Belle Foundation, collected from a 10-year-old female patient. An age- and sex-matched fibroblast cell line from a healthy individual was used as a control. Patient-derived cells were cultured in Ham’s F-12K medium (#211277022, Gibco, Thermo Fisher Scientific) supplemented with 20% FBS and 1% penicillin/streptomycin. Cells were plated at 15,000 cells/well in a clear 96-well plate to measure cellular viability using a commercially available kit (CCK8).

For measurements of mRNA transcript levels, HEK293T cells were seeded at 300,000 cells per well of a 6-well plate. Following overnight incubation to allow for proper attachment, cells were collected at 24 h using a commercial total RNA extraction kit, RNeasy (QIAGEN, 74104, Hilden, Germany). Similarly, patient-derived fibroblasts were plated at 75,000 cells/well in 6-well plates and collected after 4 days. Extracted RNA was stored at −20°C until used for RT-qPCR.

Transfection and treatment of mammalian cells

Plasmids containing either WT or mutant (R534X or R522X) NARS1 genes were transfected into HEK293T cells for overexpression studies. Aliquots of 100 ng of plasmid DNA per well of a 96-well plate were used for transfection with 0.24 μL of P3000 reagent and 0.4 μL of lipofectamine 3000 (#L3000001, Thermo Fisher Scientific) in Opti-MEM reduced serum medium (#31985070, Gibco, Thermo Fisher Scientific) as per manufacturer’s instructions. Cell media were replaced after 6 h to reduce transfection toxicity.

Cells were additionally treated with either Asn supplementation (200 mg/L), tRNA^Asn supplementation (transfection as aforementioned with 100 ng plasmid DNA/well), or the proprietary GCN2 inhibitor ALE2 (1 nM dissolved in DMSO, Alesta Therapeutics, Netherlands). Patient-derived fibroblasts were plated as described earlier and treated with either Asn supplementation or tRNA^Asn supplementation or ALE2 for cell viability assays. Media were replaced 6 h following transfection of tRNA^Asn, and cells were imaged 24 h following the initialization of treatment.

Cytotoxicity and cell viability assays

To measure cell viability or cytotoxicity, HEK293T cells were seeded at a density of 25,000 cells or 10,000 cells per well in clear or white-walled 96-well plates, respectively, and treated as described earlier. Following treatment, cells were analyzed with the CCK8 kit (CCK8, Abcam, Cambridge, UK) to measure cell viability after 24 h, following the manufacturer’s instructions. Background absorbance at 650 nm was subtracted from absorbance at 450 nm, measured using a Cytation C10 imager (BioTek Instruments, Winooski, VT, USA).

Cytotoxicity was analyzed using a commercially available kit (CytoTox-Glo, Promega Corporation, Madison, WI, USA), following manufacturer’s instructions. At 24 and 48 h following treatment, luminescence was measured at 600 nm using a Cytation C10 imager (BioTek Instruments, Winooski, VT, USA). Normalization was done using a total dead cell measurement as per manufacturer’s instructions to control for changes in cell numbers, and toxicity is reported as a luminescent ratio of live to dead cells.

Recombinant NARS1 purification and protein stability assays

Recombinant NARS1 proteins were produced as an N-terminal fusion protein in Escherichia coli (Rosetta (DE3) pLysS competent cells, Novagen) and purified by Ni-NTA immobilized metal affinity chromatography (Ni-NTA IMAC) as previously described.^12,18

Differential scanning fluorimetry was performed for recombinant WT and R534X NARS1 using the Protein Thermal Shift Dye Kit (#4461146, Applied Biosystems, Thermo Fisher Scientific) according to the manufacturer’s instructions and as previously described.¹² Briefly, recombinant NARS1 proteins were diluted to a final concentration of 1 μM per well in thermal shift buffer and SPYRO orange fluorescent dye with or without 1 μM Asn and/or 1 μM adenosine triphosphate (ATP). Plates were microfilm sealed, and protein unfolding was monitored using the QuantStudio 3 Real-Time qPCR System (#A28567, Thermo Fisher Scientific). Samples were subjected to an increasing temperature range of 25°C–96°C at a rate of 1°C per minute, and fluorescence intensity was measured every 0.017°C. Initial analysis was performed using the Protein Thermal Shift software v.1.3 (Thermo Fisher Scientific), and the data were fit to the Boltzmann equation to approximate the T_m. Statistical analysis was performed using GraphPad Prism.

Limited tryptic proteolysis of NARS1 proteins was performed as previously described.⁹ Reactions were incubated at 37°C for 5, 30, 60, or 120 min and quenched with 3× SDS dye (6% sodium dodecyl sulfate [SDS], 30% glycerol, 0.006% bromophenol blue, 125 mM Tris-HCl [pH 6.8]), followed by boiling at 95°C. Controls (C0 and C120) were incubated without trypsin for 0 or 120 min, respectively. NARS1 proteins were equilibrated with an equivalent volume of buffer before digestion.

MS

Sample preparation for MS analysis

HEK293T cells were seeded in 6-well plates at a density of 300,000 cells and grown in DMEM medium (#11995073, Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% FBS and 1% penicillin/streptomycin in a humidified incubator at 37°C with 5% CO₂. The following day, upon reaching ∼70% confluency, cells were transfected, and media were replaced, supplemented with or without 200 mg/L Asn or 1 nM ALE2. Transfected cells were left overnight to allow uptake of lipocomplexes and washed once with phosphate-buffered saline (PBS) (#14190144, Gibco, Thermo Fisher Scientific). Cells were then lifted with 0.25% Trypsin- ethylenediaminetetraacetic acid (EDTA) (#25200056, Gibco, Thermo Fisher Scientific) and centrifuged at 700 × g for 15 min at 4°C. Residual media and trypsin were removed in an additional wash with PBS and centrifuged at 5,000 × g for 5 min at 4°C. Cells were stored at −80°C until ready to be digested. HEK293T cell pellets were resuspended with 8 M urea and 2% SDS in 100 mM Tris-HCl (pH 7.4). Each sample was sonicated at an amplitude of 60% for 5 min with cycles of 30 s on/30 s off. Cell debris was pellet, and soluble protein was quantified using Pierce bicinchoninic acid (BCA) Protein Assay kit. An aliquot of 10 μg per sample was reduced with 100 mM tris (2-carboxyethyl)phosphine (TCEP) and alkylated with 200 mM Iodoacetaminde (IAA). The lysate was then digested with trypsin (Promega Corporation, Madison, WI, USA) at a 50:1, protein:enzyme ratio according to manufacturer’s recommendations. Digested peptides were desalted by C18 clean-up, lyophilized, and stored until analysis.

Tandem mass spectrometry analysis

The digested peptides were resuspended in buffer A (0.1% FA) and injected into the timsTOF Pro (Bruker Daltronics, Bremen, Germany) by nanoflow liquid chromatography using a Thermo Scientific EASY-nLC 1000 (Thermo Fisher Scientific, Massachusetts, USA). Liquid chromatography was performed using a constant flow of 300 μL/min and a 15 cm reversed-phased column with a 75 μm inner diameter filled with Reprosil C18 (PepSEP, Bruker, Germany). Mobile phase A was 0.1% formic acid and mobile phase B was 99.9% acetonitrile and 0.1% formic acid (FA). Peptide separation was carried out over 55.5 min as follows: linearly 4% A to 35% B over 50 min with an increase to 100% B over 30 s and held constant for 5 min to clean the column. Column equilibration was done prior to sample loading automatically. The trapped ion mobility spectrometry (TIMS)- time of flight (TOF) Pro was outfitted with a CaptiveSpray source (Bruker Daltronics, Bremen, Germany), operated in parallel accumulation—serial fragmentation (PASEF) mode. Trapped ion mobility separation was achieved using an accumulation time of 100 ms in the first TIMS region and ramps of the TIMS region from 0.85 to 1.30 V^.s/cm², with each ramp lasting 100 ms. MS and tandem MS (MS/MS) scans were limited to 100 to 1,700 m/z, and a polygon filter was applied to the m/z and ion mobility dimensions to select for multiple charged ions most likely to be peptide precursors. Collision energy was applied as a function of ion mobility with a linear regression using the following parameter settings: 0.85 V^.s/cm² → 27 eV, 1.30 V^.s/cm² → 45 eV. TIMS voltage was calibrated using ions from the Agilent Tune Mix (m/z 622, 922, and 1,222). Active exclusion of tandem mass spectrometry (MS/MS) scans was enabled at a setting of 0.40 min. Quadrupole isolation was set to 2 m/z for m/z less than 700, and 3.0 m/z for ions with an m/z greater than 800. A linear regression calculation was done automatically for ions between m/z 700 and 800.

Proteomics data search

MS raw data files were processed using PEAKS Studio 12.5 (Bioinformatics Solutions Inc., Ontario, Canada). The data were searched against the Homo sapiens Uniprot database. Precursor ion mass error tolerance was set to 10 ppm, and fragment ion mass error tolerance was set to 0.02 Da. Semi-specific cleavage with trypsin was selected with a maximum of two missed cleavages. A fixed modification of carbamidomethylation (+57.02 Da) on cysteine residues and variable modifications of deamidation (+0.98 Da) on Asn and glutamine, as well as oxidation (+15.99 Da) on methionine, were specified. A 1% false discovery rate (FDR) was set for the database search. After the database search, mutations were manually searched. Only mutations supported by relative fragment ion intensities >2% and having a PEAKS AScore of >15 were considered in the analysis.

Quantitative reverse-transcription PCR

For RT-qPCR, cells were collected, and RNA was extracted as described earlier using a commercial RNeasy Mini Kit (QIAGEN, Germany) as per the manufacturer’s instructions. Further cleanup of RNA was done using phenol-chloroform extraction. For input, 2 μg of RNA was used for cDNA synthesis using the high-capacity cDNA reverse transcription kit (#4368814, Thermo Fisher Scientific, USA) as per the manufacturer’s instructions. qPCR was performed using PowerTrack SYBR Green Master Mix (#A46012, Thermo Fisher Scientific) and measured with a QuantStudio3 instrument (#A28567, Thermo Fisher Scientific). Negative control reactions containing either no reverse transcriptase or no cDNA were performed in parallel to monitor DNA contamination or primer-primer interactions. Results were quantified using the ΔΔCT method and normalized to beta-actin. Primers used for the experiment are listed in Table S1.

Western blotting

HEK293T cells were grown as described earlier. 24 h after transfection, media were aspirated, and cells were washed once with PBS (#14190144, Gibco, Thermo Fisher Scientific). Cells were then lifted with 0.25% Trypsin-EDTA (#25200056, Gibco, Thermo Fisher Scientific) and centrifuged at 700 × g for 15 min at 4°C. Supernatant was aspirated, and the cell pellet was rinsed with PBS, followed by centrifuging at 5,000 × g for 5 min at 4°C. Cells were then lysed (50 mM Na₂HPO₄, 1 mM NaP₂O₇, 20 mM NaF, 2 mM ethylenediaminetetraacetic acid [EDTA], 1% Triton X-100, 1 mM dithiothreitol [DTT], 2 mM phenylmethylsulfonyl fluoride [PMSF], and 3.2 mg/mL protease inhibitor). Following centrifugation at 15,000 × g for 5 min at 4°C, the supernatant was collected, and protein concentration was determined using a BCA assay (#23227, Pierce, Thermo Fisher Scientific).

Lysates were suspended in 3× SDS buffer and separated by 12% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) at 100 V (#1658004, Bio-Rad Laboratories), followed by wet transfer onto nitrocellulose membrane (#1620115, Bio-Rad Laboratories) using a Mini Trans-Blot Electrophoretic Transfer Cell (#1703930, Bio-Rad Laboratories). Primary antibodies against the V5 tag (Abcam, #ab206566-1001) to detect NARS-V5 fusion protein, eIF2α (#9722, Cell Signaling Technology), and peIF2α^Ser51 (Abcam, #ab32157) were used at 1:1,000 dilution in bovine serum albumin (BSA)-Tris-buffered saline with Tween 20 (TBST). Specific antibodies for tubulin (T8203, Millipore Sigma), GAPDH (MAB374, Millipore Sigma), or β-III-tubulin (Abcam, #ab78078) were used as loading controls as indicated and added at a 1:1,000 dilution in BSA-TBST. Secondary antibodies IRDye 800CW Donkey anti-rabbit IgG, goat anti-mouse IgG (H + L)-HRP conjugate (#1706516, Bio-Rad), and 600CW goat anti-mouse (LI-COR) were used at 1:5,000 and 1:10,000 dilutions, respectively. Blots were imaged with a ChemiDoc MP Imaging System (#734BR5528, Bio-Rad Laboratories) and analyzed using ImageLab software.

Nuclear magnetic resonance spectroscopy

To investigate changes in extracellular levels of selected amino acids in conditioned media, 1 mL of growth media was collected from each cell culture flask after 24 h of incubation with treatment and centrifuged for 5 min at 200 × g to pellet cells, followed by a 20-min centrifugation of the supernatant at 21,000 × g to remove remaining debris. Samples were then flash frozen and stored at −80°C until freeze-dried (Labconco, Kansas City, MO, USA). The dried samples were then reconstituted using 600 μL of 50 mM phosphate buffer in 99.8% D₂O (Cambridge Isotope Laboratories, Inc., Andover, MA, USA) at pH 7.2 (uncorrected) with 50 μM 3-(tetramethysilane) propionic acid-2,2,3,3-d4 (TMSP) (Cambridge Isotope Laboratories, Inc., Andover, MA, USA) for spectral referencing. The samples were then centrifuged at 16,000 × g for 5 min at 4°C to remove any residual debris, and the supernatants were transferred to 5 mm nuclear magnetic resonance (NMR) tubes for data acquisition.

NMR experiments were conducted at 298 K using a Bruker Avance III HD 600 MHz spectrometer (Bruker, Bremen, Germany) with a 5 mm triple-resonance (¹H, ¹³C, ¹⁵N) room temperature probe. An automated data acquisition was carried out using a SampleJet autosampler and Bruker ICON-NMR software. One-dimensional (1D) ¹H-NMR spectra were collected with 32 K data points, a spectrum width of 6,578 Hz, 256 scans, and 16 dummy scans using an excitation sculpting pulse sequence for water suppression.⁷⁸ MestReNova software v.15.0.1 (Mestrelab Inc., Barcelona, Spain) and Chenomx Profiler v. 17.0.6 (Chenomx, Edmonton, AB, Canada) were used for spectral annotation. Metabolite peak integrals were used to quantify relative metabolite concentration. Relative metabolite compositions were calculated from spectral regions corresponding to Asn (2.89 ppm), aspartate (3.89 ppm), glutamine (2.46 ppm), and glutamate (2.04 ppm) and plotted for each group. Statistical analysis was performed using GraphPad Prism (GraphPad Software, San Diego, CA, USA).

Aminoacylation assays in cell extracts

Aminoacylation was assessed by measuring NARS1 activity in cultured patient-derived fibroblasts. Fibroblast lysates (cytosolic fraction) were incubated in triplicate at 37°C for 10 min in a reaction buffer containing 50 mL Tris-HCl, pH 7.5, 12 mM MgCl₂, 25 mM KCl, 1 mg/mL BSA, 0.5 mM spermine, 1 mM ATP, 0.2 mM yeast total tRNA, 1 mmol/L dithiothreitol, and 0.3 mmol/L [15N2]-Asn. The reaction was terminated using trichloroacetic acid. After washing the sample with trichloroacetic acid, ammonia was added to release the labeled amino acids from the tRNAs. [13C2, 15 N]-glycine was added as an internal standard, and the labeled amino acids were quantified by liquid chromatography MS/MS (LC-MS/MS). To investigate the effect of temperature and lower Asn availability on NARS1 aminoacylation activity, the assay was repeated after incubating the cell lysate for 10 min at 42 °C. The assay was then conducted as described with reduced Asn at 0.075 mM [15N2]-Asn.

Statistical analysis

All statistical significance was determined one-way ANOVA for multiple comparisons and corrected with Bonferroni correction or Tukey’s Honestly Significant Difference (HSD) test as indicated. Statistical significance levels are annotated using asterisks (∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, ns, not significant).

Data and code availability

Tables S1 and S2 as well as Figures S1–S4 are available online. The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD068839.

Acknowledgments

We thank Vicky Siu and Angelica Moresco for discussions and the contributing families from The Rory Belle Foundation for advice, biospecimen samples, and enthusiastic discussions and for relentlessly advocating for patients with NARS1 disease. This work was supported by grants from Natural Sciences and Engineering Research Council of Canada (04776 and 04835 to I.U.H. and 04282 to P.O.), Canada Research Chairs (232341 to P.O.), the Canadian Institutes of Health Research (165985 to P.O. and 185961 to I.U.H.), Ontario Ministry of Research and Innovation (ER-18-14-183 to I.U.H.), the Huntington Society of Canada Endowed Research Chair (to P.O.), a Rare Disease Models and Mechanism Grant (to I.U.H.), an ENRICH Doctoral Fellowship to (T.N.S.), and the Rory Belle Foundation (to I.U.H.). The NMR studies were enabled by the Biocore facility at Western University.

Author contributions

T.N.S.: conceptualization, methodology, formal analysis, investigation, writing – original draft, writing – review and editing, and visualization; J.H.K.: methodology, formal analysis, investigation, writing – original draft, and visualization; C.W.: methodology, validation, formal analysis, investigation, writing – original draft, and visualization; F.A.: methodology, formal analysis, investigation, data curation, writing – original draft, and visualization; M.I.M.: methodology, formal analysis, investigation, data curation, writing – original draft, and visualization; D.E.S.: methodology, formal analysis, investigation, data curation, writing – original draft, and visualization; R.H.: resources, writing – original draft, and funding acquisition; F.W.: data curation; J.M.S.: resources and funding acquisition; G.S.S.: methodology, formal analysis, investigation, data curation, writing – original draft, and visualization; K.H.: software, data curation, and writing – original draft; T.G.: conceptualization, resources, project administration, and writing – original draft; P.O.: conceptualization, resources, writing – original draft, writing – review and editing, and project administration; I.U.H.: conceptualization, resources, writing – original draft, writing – review and editing, supervision, project administration, and funding acquisition.

Declaration of interests

J.M.S. is the Vice President of Research & Development, Alesta Therapeutics. R.H. is the Co-Founder and Director of Research & Clinical Collaboration, The Rory Belle Foundation. I.U.H. received research funding from The Rory Belle Foundation and Alesta Therapeutics. Alesta Therapeutics owns patent “PCT/US2024/050081” describing GCN2 inhibitors including ALE2.