Functional Analysis of Missense DARS2 Variants in Siblings with Leukoencephalopathy with Brain Stem and Spinal Cord Involvement and Lactate Elevation

Abstract

Biallelic pathogenic variants in the nuclear gene DARS2 (MIM# 610956), encoding the mitochondrial enzyme aspartyl-tRNA synthetase (MT-ASPRS) cause leukoencephalopathy with Brain Stem and Spinal Cord Involvement and Lactate Elevation (LBSL) (MIM# 611105), a neurometabolic disorder characterized by progressive ataxia, spasticity, developmental arrest or regression and characteristic brain MRI findings. Most patients exhibit a slowly progressive disease course with motor deterirartion that begins in childhood or adolescence, but can also occasionaly occur in adulthood. More severe LBSL presentations with atypical brain MRI findings have been recently described. Baker’s yeast orthologue of DARS2, MSD1, is required for growth in oxidative carbon sources. A yeast with MSD1 knockout (msd1Δ) demonstrated a complete lack of oxidative growth which could be rescued by wild-type MSD1 but not MSD1 with pathogenic variants. Here we reported two siblings who exhibited developmental regression and ataxia with different age of onset and phenotypic severity. Exome sequencing revealed 2 compound heterozygous missense variants in DARS2: c.473A>T (p.Glu158Val) and c.829G>A (p.Glu277Lys); this variant combination has not been previously reported. The msd1Δ yeast transformed with plasmids expressing p.Glu259Lys, equivalent to human p.Glu277Lys, showed complete loss of oxidative growth and oxygen consumption, while the strain carrying p.Glu158Val, showed a significant reduction of oxidative growth, but a residual ability to growth was retained. Structural analysis indicated that p.Glu158Val may interfere with protein binding of tRNA^Asp, while p.Glu277Lys may impact both homodimerization and catalysis of MT-ASPRS. Our data illustrate the utility of yeast model and in silico analysis to determine pathogenicity of DARS2 variants, expand the genotypic spectrum and suggest intrafamilial variability in LBSL.

Introduction

Leukoencephalopathy with Brain Stem and Spinal Cord Involvement and Lactate Elevation (LBSL) (MIM# 611105) is caused by biallelic pathogenic variants in the nuclear gene DARS2 (MIM# 610956), which encodes the mitochondrial enzyme aspartyl-tRNA synthetase (MT-ASPRS) [1, 2]. MT-ASPRS catalyzes the production of L-aspartyl-tRNA (tRNA^Asp) [3], an important step in mitochondrial protein synthesis [4]. The reduced or absent MT-ASPRS function impairs mitochondrial translation and subsequently abnormal cellular respiration and energy production as well as mitochondrial dynamics [4].

LBSL is typically suspected in patients presenting with progressive ataxia, spasticity, developmental arrest or regression [5, 6]. Magnetic resonance imaging (MRI) studies show characteristic signal abnormalities of the white matter in the brain and spinal cord [6–8]. Magnetic resonance spectroscopy (MRS) typically also identifies lactate elevation in the affected areas [7]. Although the onset of LBSL symptoms varies from antenatal to adulthood, the most common form occurring in childhood [5, 6, 9]. The progression of the disease is variable but is typically slow, especially in childhood- and adult-onset forms [6, 10, 11]. One study showed that the time patients become wheelchair dependent correlates with age of onset of LBSL [6]. A recent study identified two additional phenotypes: a severe antenatal and neonatal form characterized by microcephaly, profound developmental delay, and intractable epilepsy with rapid progressive course leading to early death and an early infantile-onset form with variable but less severe neurological findings associated with white matter abnormalities without long tract involvement [6]. Although different studies support certain degree of genotype-phenotype correlation, some variable expressivity with similar genotypes has been observed [12, 13].

Baker’s yeast (Saccharomyces cerevisiae) has been successfully used as a model organism to study a variety of human mitochondrial disorders, including mitochondrial tRNA synthetase defects [14, 15]. Similar to human, yeast has two isoforms of aspartyl-tRNA synthetase: cytosolic and mitochondrial [14]. The yeast homologue of human DARS2, MSD1, encodes the mitochondrial isoform of aspartyl-tRNA synthetase. A yeast with MSD1 knockout (msd1Δ) is characterized by a complete lack of growth on oxidative carbon sources [6]. Expression of wild-type MSD1 in msd1Δ strain was shown to rescue this phenotype; however, expression of MSD1 harboring pathogenic variants did not restore at all or did not fully restore the oxidative phosphorylation defect thus validating the pathogenic variants [6].

Here, we report two siblings with compound heterozygous missense likely pathogenic variants in DARS2. Given the conserved biology and ability for genetic manipulation, we used yeast as a model to study the effect of novel DARS2 variants. Functional analysis and protein structure analysis confirmed the pathogenicity of the variants. Our work demonstrates the utility of functional analysis and protein modelling to study the pathogenicity of genetic variants and suggests intrafamilial variability in clinical presentation and brain imaging in patients with LBSL.

Materials and Methods

Case Presentation

Patient 1

Patient 1 is a male of Northern European ancestry who was born at 34 weeks of gestation to a 24-year-old mother. His birth weight was at 38^th percentile and birth length was at 92^nd percentile. He was transferred to the neonatal intensive care unit (NICU) and stayed for 23 days due to prematurity related concerns including intermittent requirement for oxygen supplementation for the first 5 days of life due to respiratory distress and apnea of prematurity. He required tube feedings due to poor oral intake. He was developing appropriately until 9 months of age when his parents started noticing stiffness in his legs in sitting and standing positions. He subsequently began exhibiting developmental regression and was not able to sit or stand. Brain MRI at 11 months revealed symmetric extensive abnormal T2 hyperintense signal within the periventricular white matter of the cerebral hemispheres predominantly posteriorly with mild degree of signal abnormality in the frontal lobes. The area of abnormal signal also exhibited abnormal diffusion signal. The signal abnormalities extended into the posterior limb of the internal capsule and pons. Given concerns for leukodystrophy, he underwent extensive testing including lysosomal enzymes activity, peroxisomal very long fatty acids profile and carbohydrate deficient transferrin which were normal. Lactate was normal. He subsequently underwent clinical exome sequencing, which revealed two DARS2 variants that were classified by the lab as of uncertain significance (VUS). Based on radiological findings, clinical course and molecular data he was diagnosed with presumed LBSL. He received extensive therapies and was prescribed mitochondrial cocktails containing levocarnitine, coenzyme Q10, Vitamin E, and Vitamin B6. His parents reported subjective improvement in energy levels as well as stability in walking and standing. He started walking independently at 2.5 years of age; however, his gait was unsteady. He had persistent pronation of his distal lower extremities and required orthosis. He started using utensils and was able to build a tower with blocks at 2 years of age.

By 4 years of age, he was able to walk, jump and climb. However, his gait became unstable with acute illnesses. He was able to feed himself with utensils. He could dress and undress himself with assistance. He could speak in full sentences and had ample vocabulary. He could count to 20 and knew colors and body parts. Physical examination at age 4 years revealed a non-dysmorphic child with decreased axial tone and increased appendicular tone, generalized hyperreflexia and ataxic gait with widened base and pronation of his distal lower extremities bilaterally. The pronation was thought to refelect muscle weakness in his feet.

At 4.5 years of age, he underwent follow-up brain and spine MRIs which again demonstrated bilateral periventricular white matter T2/FLAIR hyperintensities, also involving the splenium of the corpus callosum, posterior limb of the internal capsule (Figure 1A–B), and dorsal columns of the spinal cord (Figure 1D–F). Areas of abnormal diffusion signal were again noted in the parietal lobes (Figure 1G). MR spectroscopy (MRS) of the parietal lobe demonstrated an elevated N-acetylaspartate (NAA) peak (Figure 1C).

Figure 1. MRI images of the probands.

Axial brain images of Patient 1 revealed bilateral periventricular white matter T2/FLAIR hyperintensities (A, B, arrows). MRS demonstrated elevated NAA (C, arrow). Axial (D, E) and sagittal (F) spine images of Patient 1 showed elevated T2 signal over the entire spinal cord and in the dorsal columns (marked by both white and black arrows as the bright area in the posterior part of the spinal cord). Both Patient 1 (G) and Patient 2 (H) demonstrated abnormal diffusion signal in the parietal lobes (arrows).

At 5 years of age, he developed difficulties chewing solid food as well as constipation which was managed by laxatives. He also experienced a generalized tonic-clonic seizure and underwent electroencephalography (EEG) showing prominent interictal discharges in the right posterior and midline head regions which further increased in prominence during sleep. EEG background showed excessive beta frequencies. Of note, the patient did not receive antiepileptics prior to or during EEG and had no observable spells during EEG recording. He had two additional seizures and required levetiracetam for seizure control.

At 6 years of age, he demonstrated consistent developmental progress but continued exhibiting unstable gait with fatigue, illnesses, or extreme hot or cold weather. He could add and subtract but struggled with writing and tasks requiring hand skills due to muscle weakness and complains of fatigue in his hands but otherwise did well academically. His physical and neurological examination at 6 years of age was generally unchanged. His weight was at 45^th percentile, height was at 32^nd percentile, and occipitofrontal circumference (OFC) was at 73^rd percentile. His metabolic evaluation revealed normal plasma lactate at 0.8 mmol/L (reference range 0.8–2.0 mmol/L). Plasma amino acid demonstrated elevated alanine to 435 mmol/L (reference range 150–400 mmol/L).

Patient 2

Patient 2 is the younger brother of Patient 1. He was born at 38 weeks following uncomplicated pregnancy and had unremarkable neonatal course. His birth weight was at 15^th percentile and birth length was at 74^th percentile. The parents noticed that he was commonly irritable and took extended periods to console in early infancy.

He was evaluated at 7 weeks of age and noted to have increased appendicular tone with normal axial tone and reflexes. He had not gained full head control. An MRI of the brain and spine performed during this evaluation demonstrated mild diffusion signal abnormality in the parietal lobes (Figure 1H).

At 8 months of age, the parents noticed that he had increasing irritability and progressive weakness, which started after a respiratory syncytial virus pneumonia. He was started on levocarnitine. He also exhibited poor feeding and constipation leading to a hospital admission. Physical examination was notable for significant irritability, worsening appendicular hypertonia and axial hypotonia with significant head lag.

On his re-assessment at 9 months of age, he exhibited developmental regression and feeding difficulties. He was recommended to receive extensive therapies and started on coenzyme Q10 and baclofen with subsequent improvement in his hypertonia. His ophthalmological assessment showed amblyopia and strabismus. He was switched to the same mitochondrial cocktail that his sibling was receiving.

On his evaluation at 3 years of age, he demonstrated gradual but slow progress in his development. He was able to sit and crawl but could not stand or walk independently. He could reach and transfer objects. He pointed with his whole hand. He could speak 10–20 words and was starting to put words together. His constipation improved on daily use of polyethylene glycol. On physical examination, his weight was at 4^th percentile, height was at 3^rd percentile and OFC was at 54^th percentile. Neurological examination showed appendicular hypertonia, hyperreflexia, and axial hypotonia. Metabolic evaluation revealed mildly elevated plasma lactate to 2.1 mmol/L (reference range 0.8–2.0 mmol/L) and mildly elevated plasma alanine to 418 mmol/L (reference range 150–400 mmol/L).

Molecular analysis

Clinical exome sequencing in Patient 1 and targeted Sanger sequencing of the familial variants in Patient 2 were performed by GeneDx (Gaithersburg, MD). The variants were confirmed by Sanger sequencing. Variant annotation and analysis were performed using company’s custom-developed analysis tool. In silico analysis of the effect of missense variants was performed by various bioinformatic tools including Combined Annotation Dependent Depletion (CADD) [16], Rare exome variant ensemble learner (Revel) [17], VARITY [18], Sorting Intolerant From Tolerant (SIFT) [19], MutationTaster2 (MT) [20], and Functional Analysis through Hidden Markov Models (FATHMM) [21]. Variant classification was performed according to the American College of Medical Genetics and Genomics (ACMG) recommendation [22].

Yeast strains and Growth Conditions

The yeast strains used in this study were derived from W303-1B (Matα ade2-1 leu2-3, 112 ura3-1 trp1-1 his3-11, and 15 can1-100). The yeast MSD1 gene was previously cloned into the pFL38 and pFL39 vector under the control of its natural promoter. The pFL38-MSD1 was introduced in the W303–1B strain and then the endogenous MSD1 gene was disrupted, obtaining W303–1B msd1Δ/pFL38-MSD1 as previously described [6]. MSD1 was mutagenized by a PCR overlap technique to obtain mutant alleles and cloned in the pFL39 vector. Wild-type and mutant alleles each carrying an HA epitope at the 3’ end for immuno-visualization were also produced.

The empty plasmid and the plasmids carrying the wild-type or the mutant alleles, with or without the HA-tag, were introduced in the W303–1B msd1Δ/pFL38-MSD1 strain using the LiAc-ssDNA-PEG method, and transformation performed after growth in YPAD medium. The pFL38-MSD1 was lost through plasmid-shuffling on 5-fluoroorotic acid (5FOA)-containing medium thus obtaining the strains under analyses.

For all the experiments except for transformation, cells were grown in liquid SC medium (0.69% YNB without amino acids and 0.5% ammonium sulfate (Formedium^™, UK), added of 1g/l dropout mix without tryptophan [23]) in constant shaking at 28°C or in solid SC medium using 20 g/L agar for solidification (Formedium^™, UK). Media were supplemented with various carbon sources (Carlo Erba Reagents, Italy) as indicated in the results and figures. For growth analyses the strains were serially diluted, spotted, and grown at 28°C on SC medium agar plates supplemented with a fermentable carbon source, 2% glucose, or a non-fermentable carbon source, 2% glycerol, lactate or acetate.

Oxygen Consumption Assay

Mitochondrial respiratory activity in yeast was evaluated by measuring oxygen consumption, using Clark-type oxygen electrode (Oxygraph System Hansatech Instruments England) at 30°C with 1 ml of air-saturated respiration buffer (0.1 M phthalate–KOH pH 5.0, 0.5% glucose) from yeast cell suspensions cultured for 18 hr at 28°C in liquid SC medium supplemented with 0.6% glucose until exhaustion.

Protein quantification

Yeast cells were grown as for oxygen consumption assay and protein extraction was performed with the trichloroacetic acid method. Proteins were resuspended in Laemmli sample buffer pH 6.8, separated on 4–15% precast gels (Bio-Rad, USA) and Western blotted on nitrocellulose filters. Proteins were detected with primary antibodies for HA-tag (1:1000) and Por1(1:7500), followed by fluorescent secondary antibodies (anti-rat DyLight 800 1:5000, anti-mouse StarBright^™1:5000), signals were detected using Chemidoc MP Imaging System and quantified with Image Lab software (Bio-Rad, USA).

Protein Modelling of DARS2 variants

Structural analysis was based on the crystal structure of the homodimeric human DARS2 (Protein Data Bank, PDB, entry 4AH6). Aspartyl-adenosine-5’-monophosphate and tRNA^Asp ligands were added to this structure employing same binding poses as in their crystallographic complexes with the homologous aspartyl-tRNA synthetase from T. thermophilus (PDB entries 1G51 and 1EFW).

Results

Next generation sequencing identified novel variants in DARS2

Clinical exome sequencing identified two heterozygous missense variants in DARS2 (NM_018122.4): paternally-inherited variant designated c.473A>T (p.Glu158Val), which was reported once in a compound heterozygous state with the recurrent c.492+2T>C variant in a 45-year-old patient [24], and novel maternally-inherited variant designated as c.829G>A (p.Glu277Lys). The p.Glu277Lys variant was reported in gnomAD (frequency 1/251,350) while the p.Glu158Val was not found in the same database [25].

In silico analysis and protein modelling

Both variants are highly conserved among many species (Figure 2A). In silico analysis supported deleterious effect of protein structure and function (Table 1). Structural analysis based on the crystal structure of the homodimeric human DARS2 allowed to understand how both variants can affect the protein function (Figure 2B). The p.Glu158Val variant replaces a glutamic acid residue conserved from mammals to insects that is part of the tRNA^Asp binding region of MT-ASPRS. Glu158 has a functional importance because it presents electrostatic interactions with two cationic residues, Lys160 and Lys164, and together determine the electrostatic properties of the sites and its binding to the negatively charged tRNA^Asp. Thus, the replacement of the anionic glutamate with a neutral valine is expected to introduce defects in the interactions between MT-ASPRS and the cognate tRNA^Asp.

Figure 2. MT-ASPRS protein sequence alignment and structural analysis.

A. Multiple sequence alignment highlights the conservation of the sites affected by the p.Glu158Val and p.Glu277Lys variants (identical and similar residues with 50% conservation or above are shadowed in black and gray, respectively). B. MT-ASPRS structural analysis. Glu158 and Glu277 (sticks with meshes) are indicated in the crystal structure of the human DARS2 homodimer (PDB 4AH6; protein monomers in distinct colours). The enlarged views show the residues interacting with Glu158 and Glu277. To highlight the functional importance of these sites, aspartyl-adenosine-5’-monophosphate and tRNA^Asp, available from their crystallographic complexes with the aspartyl-tRNA synthetase from T. thermophilus (PDB 1G51 and 1EFW) were added to the human DARS2 homologue employing analogous binding poses. Amino acids are numbered according to NCBI sequence NP_060592.2.

Table 1.

In silico prediction and ACMG classification of the two DARS2 variants.

Empty Cell	c.473A>T (p.Glu158Val)	c.829G>A (p.Glu277Lys)
CADD	32	32
Revel	Deleterious (low) (0.69)	Deleterious (0.92)
Varity	Deleterious (0.76)	Damaging (0.99)
SIFT	Tolerated (0.15)	Damaging (0)
MT	Deleterious (1)	Deleterious (1)
FATHMM	Deleterious (low) (−1.7)	Deleterious (−2.83)
ACMG Criteria	PS3 (functional studies) PM2 (population data) PP2 (low rate of benign missense variants in DARS2) PP3 (in silico predictions) PP4 (phenotype)	PS3 (functional studies) PM2 (population data) PP2 (low rate of benign missense variants in DARS2) PP3 (in silico predictions) PP4 (phenotype)
Interpretation	Likely Pathogenic	Likely Pathogenic

The p.Glu277Lys variant hits a highly conserved glutamic acid residue, which contributes to the region of MT-ASPRS engaged in homodimerization and is also important for the protein fold near the catalytic site. In fact, Glu277 presents multiple interactions, most of which electrostatic, with surrounding residues near the functional ligand aspartyl-adenosine-5’-monophosphate. It can be envisioned that the replacement of the anionic glutamate with a cationic lysine introduces severe structural alterations impairing both homodimerization and catalysis of MT-ASPRS.

Functional characterization of DARS2 variants using yeast model system

To validate the mutations identified in DARS2, Baker’s yeast S. cerevisiae was used as a model taking advantage of the presence of the ortholog gene MSD1. As shown by protein alignment (Figure 3A), the human residue p.Glu277, equivalent to p.Glu259 in yeast, is conserved from yeast to human. To study the effect of the alleged pathological amino acid substitution p.Glu277Lys, the corresponding yeast MSD1 codon was directly mutated, producing the mutant allele msd1^E259K (Figure 3B). The human amino acid residue Glu158, equivalent to Gln137 in yeast, is not conserved in yeast but lies in a conserved stretch (Figure 3A). For this reason, we have created both the so-called humanized version and the putative pathological allele to compare their effect. The humanized allele was created by replacing the yeast amino acid with the human wild-type amino acid, thus obtaining the msd1^hQ137E allele, while the mutant allele was created by the substitution of the yeast amino acid with the mutant version, thus obtaining the msd1^Q137V allele (Figure 3B).

Figure 3. Functional studies of DARS2 variants using yeast model system.

A. Partial alignment (Clustal Omega) of the human DARS2 and the yeast Msd1 proteins. The residues of interest are highlighted in green for conserved amino acids and in magenta for non-conserved. Amino acids are indicated as conserved (*), with strongly similar properties (:) or with weakly similar properties (.); B. Amino acid changes found in the patients and the corresponding humanized and mutant yeast alleles; C. Growth test: the msd1Δ strain harboring wild-type, humanized, mutant alleles or the empty vector were serially diluted and spotted on SC agar plates supplemented with the fermentable carbon source: glucose (SC-Glucose) or the non-fermentable carbon source: acetate (SC-Acetate), glycerol (SC-Glycerol) and lactate (SC-Lactate) and incubated at 28 °C; D. Respiratory activity: cells were grown at 28°C in SC medium supplemented with 0.6% glucose. Values are represented as the mean of at least three values ±SD. Green bars indicate wild type and humanized strains; blue bars indicate strains carrying the alleged pathological mutations; red bar indicates the null mutant strain. Oxygen consumption is expressed in nmol O₂/mg/min. Statistical analysis was performed using a two-tail unpaired Student’s t test: *** p < 0.001; ** p< 0,01; ns: not significant; E. Representative Western blot on total protein extract using an anti-HA antibody visualizing Msd1 protein, and anti-Por1 as loading control. The quantification was performed on two independent blots using Image Lab Software (Bio-Rad).

To assess the impact of the variants on mitochondrial function, we evaluated semi-quantitatively the strains’ growth capacity through spot assay analysis on different non-fermentable (or oxidative) carbon sources, and on glucose as control. MSD1-knock out yeast (msd1Δ) transformed with the empty vector exhibited complete inability to grow on the media containing non-fermentable carbon sources such as acetate, glycerol and lactate; whereas the re-expression of the wild-type MSD1 or the humanized allele msd1^hQ137E rescued this defect (Figure 3C). The expression of the msd1^E259K allele, carrying mutation p.Glu259Lys, equivalent to human p.Glu277Lys, did not lead to rescue of the oxidative defect as the strain showed a complete inability to grow on media containing non-fermentable carbon sources, similar to msd1Δ strain Figure 3C). The strain carrying p.Gln137Val, equivalent to human p.Glu158Val, showed a significant reduction of oxidative growth, but a residual ability to growth was retained. These findings indicate that both variants are deleterious in yeast thus validating the human variants as pathogenic.

To further investigate the effect of the variants on oxidative phosphorylation, we assessed respiratory activity by measuring the oxygen consumption rate. Both variants significantly reduced the respiratory rate compared to the wild-type or humanized strain, but to different extents. The strain carrying p.Glu259Lys showed a reduction of oxygen consumption rate (1.3±0.2 nmoliO₂/mg/min), comparable to the msd1Δ null mutant (1.4±0.5 nmoliO₂/mg/min), suggesting a complete loss of function of the protein. Whereas the p.Gln137Val variant caused a severe reduction on oxygen consumption rate (4.18±1.5 nmoliO₂/mg/min) compared to the humanized strain (26.3±2.1 nmoliO₂/mg/min), but a minimal residual respiratory activity was retained suggesting that this variant does not lead to a complete loss of function of the Msd1 protein (Figure 3D).

To evaluate whether the variants affected protein stability, we measured the steady-state level of the Msd1 proteins by Western blot analysis taking advantage of the strains expressing HA-tagged recombinant variants. To exclude that the HA-tag could interfere with protein function, we verified the complementation ability of the wild-type tagged allele of the respiratory growth phenotype of the null mutant (data not shown). No significant differences in the quantity of Msd1 protein was detected between mutant strains and wild-type or humanized strains, suggesting that both the variants do not interfere with protein stability (Figure 3E).

Discussion

Here, we describe two patients with LBSL who harbor compound heterozygous missense variants in DARS2, one of which is novel and the other has previously been described in 1 patient, but the combination of these variants has not been previously reported. Both p.Glu158Val and p.Glu277Lys affect conserved and functional regions of MT-ASPRS (Figure 2A). It should be noted that the conservation of Glu158 spans from mammals to insects and amino acid substitutions can occur in inferior organisms perhaps to accommodate amino acid substitutions determining the binding of different species-specific cognate tRNA^Asp molecules. On the other hand, Glu277 does not tolerate changes down to bacteria, highlighting the critical structural and functional importance of this residue.

Functional studies using yeast as model organism showed a significant oxidative phosphorylation defect for both variants, with the p.Glu259Lys variant (equivalent to human p.Glu277Lys) exhibiting a more severe impact respect to p.Gln137Val (equivalent to human p.Glu158Val), consistent with protein modelling prediction. In particular, the strain carrying the p.Glu259Lys variant showed a complete inability to grow on non-fermentable carbon sources and a completely abolished respiratory activity, like the null mutant; whereas the strain carrying the p.Gln137Val showed a strong but not total reduction of both oxidative growth and respiration. Given the residual ability to grow in oxidative carbon source of p.Glu158Val, it is likely that p.Glu158Val exhibits residual protein function. The vast majority of reported LBSL patients with the slowly progressive phenotype are compound heterozygotes, with one mild (hypomorphic) allele, which is an intron 2 leaky splicing variant in 95% of cases, ensuring adequate residual MT-ASPRS function [5, 6]. Furthermore, many patients with the recently described severe LBSL phenotype lacked the leaky splice site variant and had missense variants with more disruptive nature based on in silico analysis, further supporting correlation between residual enzyme activity and clinical severity [6]. Data obtained in yeast strongly suggest that DARS2^Glu277Lys is an amorphous allele while DARS2^Gln137Val is hypomorphic and the severity of the phenotype depends on the residual activity of the latter.

An association between the severity of the impairment of mitochondrial function in yeast and the severity of the clinical phenotype of the patients was previously reported [15, 26]. The DARS2 variants causing complete absence of MT-ASPRS were then expected to more significanly impacting yeast mitochondrial function and causing a severe oxidative growth defect. Our yeast studies confirmed that the two siblings have a severe likely pathogenic variant and a hypomophic variant explaining their relatively slowly progressive course. However, a recent study modelling novel DARS2 variants did not demonstrate perfect correlation of yeast and human phenotypes [6]. Additional genetic and environmental factors are thought to be contributing to the clinical phenotype. For example, in patients with pathogenic variants in YARS2, encoding for the mitochondrial tyrosyl-tRNA synthetase, it has been hypothesized that mtDNA haplotype background may influence phenotypic expression [27].

The patients with slowly progressive LBSL course exhibit distinct MRI pattern including cerebral white matter signal abnormalities with long tract involvement [5]. As discussed in Introduction, a recent study reported two additional and more severe phenotypes based on clinical presentation and brain imaging [6]. Based on radiological findings and clinical course, the two siblings in this report are classified as having the classic form of LBSL. The onset of regression is earlier in Patient 2 compared to Patient 1. Abnormal muscle tone and lack of full head control were noted in Patient 2 since 7 weeks of life while Patient 1 developed normally until 9 months of age. Patient 1 was able to walk independently by the age of 2.5 years, while patient 2 has not achieved independent standing or walking at 3 years of age. Patient 2 required baclofen for his more severe hypertonia in lower extremities. The difference in clinical pictures of the two probands may suggest the presence of additional genetic or environmental factors that play roles in disease onset and progression.

Interestingly, Patient 2 exhibited regression after a viral infection. It is possible that infection or stress may induce disease onset or accelerate disease progression [28]. This phenomenon is well described in other inherited leukodystrophies, including childhood ataxia with central nervous system hypomyelination/vanishing white matter (CACH/VWM) caused by pathogenic variants in the genes encoding the five subunits of the eukaryotic translation initiation factor 2B (eIF2B) [29]. The rapid progression in the setting of infection or physical stress is explained by its role in integrated stress response (ISR). In LBSL, recent study showed that 75% reported episodic worsening of ataxia or motor symptoms associated with a fever, illness or minor head injury [28]. The mechanism underlying this phenomenon has yet to be determined. Recent study observed diminished aminoacylation activity in fibroblasts from patients with leucyl-tRNA synthetase and beta subunit of phenylalanyl-tRNA synthetase deficiencies in the setting of increased temperature to 38.5 and 40 Celsius [30]. It is possible that increased body temperature may decrease MT-ASPRS activity and exacerbate neurological deterioration in LBSL patients but future in vitro studies are needed to confirm this speculation. In addition, it is known that catabolic stresses, including acute febrile illnesses, trigger symptoms in patients with mitochondrial disorders [31]. This phenomenon is likely due to increased reactive oxygen species produced by diseased mitochondria and/or energy deficit. The defect in MT-ASPRS activity impairs mitochondrial translation and may lead to decreased ability to cope with increased metabolic demand.

Our study is focused on a single family with two affected siblings and therefore the segregarion analysis of the two varinats is limited. Furthermore, the potential genotype-phenotype correlation related to these variants is based on clinical findings of the 2 siblings and a large cohort of patients is needed to validate this observation. Long-term follow-up is also needed to determine the differences in the clinical course of the 2 siblings and confirm whether the intrafamilial variability holds true in the future. Furthermore, no follow up brain imaging was performed for patient 2 to demonstrate the development of typical LBSL neuroimaging findings and compare them to those found in his sibling.

In conclusion, we report two patients from a family with LBSL who were diagnosed based on exome sequencing and found to have 2 novel variants in DARS2. Based on in silico prediction, protein structure analysis, and functional studies performed in yeast, we were able to reclassify these two variants as likely pathogenic. Our study highlights the utility of yeast model as a functional tool to study variants in genes related to mitochondrial disorders, however studies in larger cohorts are needed to establish the association of functional defects in yeast and clinical severity. Our study also expands the genotypic spectrum and suggests the presence of intrafamilial variability in LBSL although carefully analyzing additional affected siblings is required to validate this observation.