Clinical, biochemical, and genetic features associated with VARS2‐related mitochondrial disease

Abstract In recent years, an increasing number of mitochondrial disorders have been associated with mutations in mitochondrial aminoacyl‐tRNA synthetases (mt‐aaRSs), which are key enzymes of mitochondrial protein synthesis. Bi‐allelic functional variants in VARS2, encoding the mitochondrial valyl tRNA‐synthetase, were first reported in a patient with psychomotor delay and epilepsia partialis continua associated with an oxidative phosphorylation (OXPHOS) Complex I defect, before being described in a patient with a neonatal form of encephalocardiomyopathy. Here we provide a detailed genetic, clinical, and biochemical description of 13 patients, from nine unrelated families, harboring VARS2 mutations. All patients except one, who manifested with a less severe disease course, presented at birth exhibiting severe encephalomyopathy and cardiomyopathy. Features included hypotonia, psychomotor delay, seizures, feeding difficulty, abnormal cranial MRI, and elevated lactate. The biochemical phenotype comprised a combined Complex I and Complex IV OXPHOS defect in muscle, with patient fibroblasts displaying normal OXPHOS activity. Homology modeling supported the pathogenicity of VARS2 missense variants. The detailed description of this cohort further delineates our understanding of the clinical presentation associated with pathogenic VARS2 variants and we recommend that this gene should be considered in early‐onset mitochondrial encephalomyopathies or encephalocardiomyopathies.


INTRODUCTION
Mitochondria are defined as the powerhouses of the cell, since they produce a usable energy source, adenosine triphosphate (ATP), through oxidative phosphorylation system (OXPHOS). This process is carried out by the mitochondrial respiratory chain (MRC), and the ATP synthase, which are embedded in the inner mitochondrial membrane (MIM). The MRC is composed of four multi-subunit complexes (CI to CIV) and two mobile electron carriers (ubiquinone and cytochrome c) that produce a proton gradient across the MIM, which is then used by Complex V (FoF1 ATP synthase) to produce ATP [Lightowlers, Taylor, & Turnbull, 2015;Smeitink, Zeviani, Turnbull, & Jacobs, 2006].
OXPHOS proteins are uniquely under the dual genetic control of two genomes and as such, mitochondrial disorders can be caused by defects in either the mitochondrial DNA (mtDNA) or nuclear DNA [Spinazzola & Zeviani, 2009]. Mutations in several genes responsible for defects of mitochondrial protein synthesis, affecting either mtDNA or nuclear encoded genes, have been reported to cause a wide range of mitochondrial syndromes [Ghezzi & Zeviani, 2012;Rotig, 2011]. Mitochondrial aminoacyl-tRNA synthetases (mt-aaRSs) are key enzymes in mitochondrial protein synthesis since they catalyze the specific attachment of amino acids to their cognate tRNAs. Recently, an increasing number of mitochondrial disorders have been associated with variants in mt-aaRSs genes [Coughlin et al., 2015;Diodato et al., 2014ba,a;Konovalova & Tyynismaa, 2013;Simon et al., 2015]. However, for some mt-aaRSs (i.e., VARS2, MIM# 615917), very few individual patients have been described and the associated phenotype is thus poorly defined.
A homozygous missense pathogenic variant (c.1100C > T, p.Thr367Ile; NM_001167734.1) in VARS2, the gene encoding the mitochondrial valyl tRNA-synthetase, was described in a patient with a clinical picture characterized by psychomotor delay and epilepsia partialis continua associated with a mitochondrial Complex I defect [Diodato et al., 2014ba,b]. Additionally, compound heterozygous variants in VARS2 (the previously reported c.1100C > T, p.Thr367Ile plus a c.601C > T, p.Arg201Trp variant) were reported in a patient with a neonatal form of encephalocardiomyopathy [Baertling et al., 2017]. Pathogenic variants in VARS2 were also described in two patients included in large whole exome sequencing (WES) studies although detailed clinical and/or biochemical characterization was not provided [Pronicka et al., 2016;Taylor et al., 2014]. Here we report a detailed clinical and molecular description of the patients (and affected siblings) reported in the previous WES studies and describe 11 additional patients from unrelated families presenting with a severe infantile mitochondrial disorder associated with VARS2 pathogenic variants, further defining the clinical and biochemical features associated with VARS2 defects.

Standard protocol approvals and patient consent
Informed consent to participate in the study was obtained from all affected individuals or their parents in case of study participants under the age of consent. The study was approved by the Ethical Commit-

Molecular studies
P1 and P2 received clinical WES prior to enrollment in the Mount Sinai study to further investigate mt-aaRSs disorders. WES for P1 was completed at Ambry Genetics (Viejo, CA). Genomic DNA from the proband, mother, and father was isolated from peripheral blood samples, and WES was completed on the family trio. DNA samples were prepared   The WES also included mitochondrial genome screening. The mitochondrial genome was amplified by long-range PCR and subjected to paired-end library construction. The mean depth of coverage for mitochondrial sequencing was >25,000×.
WES on genomic DNA from affected individual P3 was performed at the Institute of Human Genetics (Munich, Germany) using the Sure-Select Human All Exon 50 Mb kit (Agilent, Santa Clara, CA, USA) for insolution enrichment followed by sequencing as 75 bp paired-end runs on a HiSeq2500 (Illumina) as described previously [Haack et al., 2012;Kremer et al., 2016]. This achieved an average of 143-fold coverage with 97.7% of the exome covered at least 20-fold.
P4 was investigated by a targeted NGS approach using a customized gene panel (TruSeq Custom Amplicon; Illumina, San Diego, CA, USA) containing genes associated with mitochondrial disorders, according to the procedure recently reported [Legati et al., 2016].
P8 and P10 had clinical WES as previously described [Yang et al., 2014]. Analysis of variants followed the guidelines of the American College of Medical Genetics and Genomics [Richards et al., 2015].
All the novel sequence variants identified in VARS2 have been submitted to a public database (https://www.lovd.nl/VARS2).

Histopathological studies
Quadriceps muscle biopsies were obtained from subjects P2, P3, P4, P5, and P6. Transversely orientated, frozen muscle sections (10 m) were subjected to standard histological and histochemical procedures including sequential cytochrome c oxidase (COX)/succinate dehydrogenase (SDH) histochemistry. Electron microscopy of P2 muscle was performed according to standard procedures.

Immunoblotting
Lymphoblast samples from P1 and two control lines (C1 and C2) were processed for immunoblotting as previously described [Webb et al., 2015]. The membrane was probed with mouse anti-OXPHOS cocktail at 1:1,000 (MitoSciences, ab110411, Eugene, OR, USA) and with rabbit anti-GAPDH at 1:5,000 (Sigma, G9545, St. Louis, MO, USA) used as a loading control. Antibodies were visualized using IRDye 800CW or IRDye 680RD secondary antibodies and the Odyssey Infrared Imaging System (LI-COR Biosciences, Lincoln, NE, USA).
Total protein was extracted from fibroblasts and muscle tissue of P5 and fibroblasts of P4, separated on 12% SDS-PAGE gels, transferred to PVDF membranes and analyzed by immunoblotting using

Homology modeling
Homology modeling of the mitochondrial human Valine-tRNA ligase (VARS2, NCBI: NP_001161206.1) in the amino acid interval 130-1,086 was made employing as the template the Protein Data Bank (PDB) structure 1IVS (representing the Valine-tRNA ligase from Thermus thermophilus, which shares 37% amino acid identity with human VARS2). All side chain atoms in the template structure were deleted.
Then, the residues of this backbone-only structure were renamed and renumbered (in PDB format) to the corresponding amino acids in the human VARS2 according to the pairwise sequence alignment shown in the Supp. File S1. All side chains were re-built and amino acid insertions added using MODELLER (v. 9v17) [Sali et al., 1993]  Sequence logos were calculated on the VARS2 multiple sequence alignment as displayed in the MiSynPat database (misynpat.org) including all organisms. Sequence logos were generated using WebLogo [Crooks, Hon, Chandonia, & Brenner, 2004].

Case reports
A short clinical description of the patients with biallelic VARS2 variants is reported below; pedigrees are shown in Figure 1, whereas the main clinical and laboratory findings are reported in Table 1.  Figure 1). Family history is notable for consanguinity; parents are second cousins. She presented with hypotonia and poor suck from birth, followed by developmental delay, poor coordination, dystonic movements and ataxia. At 3.5 years, she was unable to sit, crawl, or walk and her language was limited to fewer than 30 words. At 3 years and 8 months, she presented with seizures, which included generalized tonic-clonic, focal, and myoclonic types. MRI completed at 11.4 months of age was unrevealing; there were no acute intracranial findings or abnormalities of the orbits or brainstem identified. Evaluation of the parenchyma was limited due to incomplete myelination at this age; the myelination pattern was read as within normal limits for age. The second MRI, performed at 3 years and 6 months of age, revealed T2/FLAIR hyperintensity in the periventricular white matter and cerebellar hemispheres bilaterally, and diffuse cerebral volume loss. MR spectroscopy showed a small doublet peak at 1.3 ppm in voxels over the left parietal white matter and frontal horns, suggesting lactate. An additional MRI, performed at 4 years 9 months revealed more advanced ill-defined prominent T2/FLAIR hyperintensity in the periventricular white matter as well as multifocal areas of abnormal T2/FLAIR hyperintensity in the cortex and subcortical white matter.
P2 was a Caucasian female born at 39 weeks gestation by Cesarean section due to breech presentation (family 2; Figure 1). Apgar scores were 7 and 9 at 1 and 5 min, respectively. Shortly after birth, the newborn was transferred to the neonatal intensive care unit for evaluation of stridor. Nasal endoscopy revealed bilateral vocal cord paresis, and the patient was diagnosed with central and obstructive sleep apnea.
A tracheostomy was performed for airway protection at 2 months of age. Other symptoms/signs were hypotonia, hyporeflexia, exaggerated startle, staring episodes, and congenital hip dislocation. Echocardiogram completed at 2 months of age revealed moderate to severe biventricular hypertrophy with normal ventricular function and moderate pericardial effusion. Dysmorphic features noted on examination included microcephaly (occipitofrontal circumference [OFC] on 2 nd centile), a round, full face, prominent eyes with shallow orbits, a small nose with flat nasal root, a small mouth with down-turned corners, micrognathia, and inverted nipples. Initial brain MRI completed at 12 days of age was unremarkable, but a second brain MRI performed at 3 months revealed moderate to severe diffuse cerebral and cerebellar atrophy, scattered areas of cortical restricted diffusion, most pronounced around the Sylvian fissures bilaterally, as well as subtle thalamic restricted diffusion bilaterally. MR spectra showed large lactate peaks in the deep gray matter and in the subcortical white Clinical examination showed poor somatic growth and microcephaly (OFC < 3rd centile) and a severe neurological impairment, characterized by absence of head control, response to sounds and language, poor eye contact, hyperreflexia, and hypertonia. Visual evoked potential showed central conduction abnormalities, fundus oculi was normal.
The child died at 5 months, because of cardiac arrest.
P5 was born by normal vaginal delivery, at full term (family 5; tion. In addition, the cranial MRI also revealed a symmetrical increased T2 signal in the peri-trigonal white matter (Figure 2). Recent echocardiography has revealed mild concentric ventricular hypertrophy. An older sister died at the age of 21 months following pneumonia. She was also floppy and had restricted eye movements with bilateral ptosis. A diagnosis of congenital myasthenia had been considered but she died before this could be definitively investigated.
P6 was born at term by spontaneous delivery from Polish unrelated parents with a positive genetic history (family 6; Figure 1). Their older daughter presented at birth with hypotonia, laryngeal stridor, respiratory failure, he rapidly developed hypertrophic cardiomyopathy, impaired contractility and died at age of 32 following a cardiac arrest; massive cardiac hypertrophy and fatty liver accumulation were found at autopsy (family 6; Figure 1). The birth weight of P6 was 3,420 g and Apgar score was 10 at 1 min.

P11 was the first child of a consanguineous healthy couple of
Afghan origin (family 9; Figure 1). She was born at term (birth weight 3,300 g, length 51 cm), and had no neonatal problems. She was referred for investigations at the age of 6 months because of poor head control. On examination, she was noted to be hypotonic, microcephalic, and developmentally delayed. Nystagmus was noticed by the age of 1 year and at the age of 1 year 3 months a brain MRI showed cerebellar atrophy, slight thinning of the posterior corpus callosum and pathological FLAIR hyperintense signal in the thalami. At the age of 2 years, when her epileptic seizures started, she was able to say a few words and turn over, but never learned to crawl. She developed intractable multifocal epilepsy with prolonged seizures. A repeat MRI at the age of 2 years 4 months showed progression of the cerebellar atrophy, involving both cerebellar hemispheres and vermis, and FLAIR/T2 hyperintense signal in dentate nuclei and thalami (Figure 2).
Her disease progressed with regression of skills and feeding difficulties evident by the age of 4 years. She died of pneumonia at the age of 7 years.
P12, sister of patient 11, the second child of the family, was a term baby (birth weight 3,450 g/length 49 cm/OFC 35 cm, Apgar score 9 at 1 min) (family 9; Figure 1). At the age of 6 months, when first investigated for delayed motor development, she was hypotonic with poor head control and microcephaly (−2 SD). Brain MRI at the age of 1 year 1 month showed cerebellar atrophy and slightly thinning of the posterior corpus callosum (Figure 2). A 1H-MRS from the thalamus showed no lactate peak and EEG showed no epileptiform discharges. Epileptic seizures started with status epilepticus at the age of 2 years 1 month and EEG revealed multifocal spikes, spike-slow waves and general disturbance. At that age she was alert, able to say few words, able to roll over and catch toys. By the age of 4 years her disease had progressed, skills had regressed, she had developed limb spasticity and needed a gastrostomy for feeding. Epilepsy was intractable. She died of pneumonia at the age of 8 years.
P13, brother of patients 11 and 12, the third child of the family, was diagnosed with VARS2 mutations antenatally (family 9; Figure 1). He was born at term (birth weight 3,810 g, length 52.5 cm/OFC 34.0 cm (−1 SD), Apgar score 9 at 1 min). Neonatal brain MRI showed unilateral cerebellar hypoplasia. No neonatal problems were noticed. At 5 months of age, his head growth was slowed to −3 SD and he was hypotonic without any signs of cardiomyopathy in heart ultrasound.

Morphological studies
Histochemistry of P3 and P4 muscle samples was normal, whereas mild, unspecific myopathic changes in P5 and signs of neurogenic atrophy in P6 were reported (not shown). In P5, COX-deficient/SDHpositive fibers were evident (not shown).
Electron microscopy of P2 muscle disclosed relatively normal myofibrillar architecture, myofiber nuclei, sarcolemmal membranes, and basement membranes; there was abnormal/excessive accumulation of glycogen (non-membrane-bound) both intermyofibrillary and especially subsarcolemmal, which has a variably granular appearance to very focally filamentous appearance. The mitochondria appeared relatively decreased in number (sparse overall) and size, with mostly unremarkable morphology although some are distorted or atypical (with atypical cristae) (not shown).

Autopsy findings
For P2, post-mortem examination of the heart revealed severe biventricular hypertrophy and mildly asymmetrical hypertrophy of the ventricular septum; microscopic sections showed prominent and diffuse myocyte vacuolar degeneration. Neuropathological evaluation revealed a subacute and acute necrotizing encephalopathy consistent with Leigh syndrome with lesions evident in the thalamus and cortex, having typical histopathology of vacuolated neuropil with gliosis and prominent vasculature. Cerebellar cortical degeneration was also noted; the substantia nigra appeared relatively spared (not shown).
In P6, post-mortem examination showed hydrothorax and hydropericardium, and left ventricle hypertrophy with immature cardiac cells.

Biochemical studies
The activities of all the MRC complexes were within control range in P5 and P7 fibroblasts, whereas a partial reduction of CIV/CS (69% of mean control value) was observed in P4 fibroblasts. In these cells respiratory capacity, assessed by Seahorse micro-scale oxygraphy, was normal in standard, glucose medium but was decreased in galactose-medium, a condition that forces cells to use OXPHOS to produce ATP rather than glycolysis ( Figure 3A).
MRC complex activities in muscle samples disclosed a combined CI and CIV reduction in P5 (60% of mean control values), a CIV reduction in P3 (17% of mean control) and P6 (25% of mean control), whereas were normal in P2 and P4 muscles.

WES and Sanger validation
Details of the identified VARS2 variants are provided in In P6, older affected brother of P7, the same p.Thr367Ile and p.Arg497His mutations were identified by Sanger sequencing.
Clinical WES was conducted on P8, which reported the child was homozygous for c.1258G > A (p.Ala420Thr). Sanger sequencing confirmed the presence of the mutation in the child and in the proband's sibling P9, as well as the heterozygous state of the parents. Patient P10 also e clinical WES: although he was reported to be unrelated to family of P8-P9, the same homozygous c.1258G > A (p.Ala420Thr) variant was identified.
Clinical WES was also conducted on P12, and it detected the same homozygous c.1100C > T (p.Thr367Ile) variant in VARS2 as in P1.
Sanger sequencing confirmed the presence of the mutation in the proband and in her affected siblings (P11 and P13), as well as their parents' heterozygosity.

Functional studies
In order to assess the effect of the identified VARS2 variants on protein levels, immunoblotting was performed using available patient samples (P1 lymphoblasts, P4 fibroblasts, and P5 muscle). NDUFB8

F I G U R E 3 Functional studies. A:
Micro-oxygraphy performed in P4 and control fibroblasts cultured in galactose medium. Y-axis values correspond to the maximal respiration rate, expressed as pMolesO 2 /min/cell. Data are represented as mean ± SD. Two-tail, paired t-test was applied for statistical significance (***P < 0.001). B: Western blot analysis of P1 lymphoblasts using anti-OXPHOS cocktail (ATP5A, UQCRC2, SDHB, MTCOII, NDUFB8) and anti-GAPDH antibodies. C: Western blot analysis of P4 fibroblasts using antibodies against VARS2, mitochondrial encoded Complex IV subunit 1 (MTCOI), Complex II subunit A (SDHA), and GAPDH, the latter being used as a loading control. D: Western blot analysis of P5 muscle sample using antibodies against VARS2, Complex I subunit (NDUFB8), Complex II subunit (SDHA), Complex III core protein II (UQCRC2), Complex IV subunits (MTCOI, MTCOII), Complex V subunit (ATP5A) and alpha-tubulin, the latter being used as a loading control. E: De novo metabolic labeling in P5 and control fibroblasts. Separated on 15% PAA gel. Coomassie blue stain as loading control. F: Western blot analysis of P5 fibroblasts using antibodies against VARS2, NDUFB8, Complex IV subunits (MTCOII, COXIV), VDAC and beta-actin, the last two being used as loading controls and MTCOII protein levels were reduced in P1 lymphoblasts compared with control samples ( Figure 3B). A decrease in steady state level of VARS2 protein was observed in P4 fibroblasts compared with control lines ( Figure 3C); furthermore, a reduction in MTCOI levels was present in P4, whereas SDHA levels were unaffected ( Figure 3C). P5 muscle showed normal VARS2 steady state levels but decreased amounts of Complex I (NDUFB8) and Complex IV (MTCOI and MTCOII) subunits compared with a control muscle ( Figure 3D).
Next we determined the consequence of the VARS2 mutations on mitochondrial translation by performing de novo metabolic labeling in patient and control fibroblasts. Data analysis in P5 fibroblasts compared with aged-matched control cells indicated that mutant VARS2 caused no significant changes in the rate of mitochondrial translation ( Figure 3E). Immunoblotting analysis was performed on the same fibroblasts and also showed no differences in VARS2 levels or steady-state levels of Complexes I and IV subunit ( Figure 3F), consistent with the normal mitochondrial protein synthesis and reports of other mutated mitochondrial aa-tRNA synthetases, where the consequential biochemical defect is usually expressed in muscle but not in cultured fibroblasts [Almalki et al., 2014].

DISCUSSION
Mitochondrial diseases due to defects in mt-aaRS are emerging as an important category of mitochondrial disorders. Despite reporting of a strict genotype-phenotype association for many mt-aaRS, in leukodystrophy with ovarian failure [Dallabona et al., 2014]. Other examples are the different clinical phenotypes caused by FARS2 [Almalki et al., 2014;Elo et al., 2012], LARS2 mutations [Pierce et al., 2013;Riley et al., 2016], or MARS2 mutations [Bayat et al., 2012;Webb et al., 2015]. It remains poorly understood why genetic defects in mt-  et al., 2014ba,b;Taylor et al., 2014], and the biochemical phenotype caused by mt-aaRS mutations [Diodato et al., 2014ba,b], which is frequently normal in fibroblasts cell lines.
The mutant VARS2 allele c.1100C > T (NM 001167734.1) is recurrent in our population, either in homozygosity (P1; P11-P13) or in compound heterozygosity (P2, P4, P5, P6-7). The c.1100C > T allele was present in homozygosity in the first patient of Italian ancestry [Diodato et al., 2014ba,b], and was also present in a Greek patient [Baertling et al., 2017] Five out of the six missense mutations affect residues located in regions important for the interaction of VARS2 with the cognate tRNA (p.Thr367Ile, p.Ala379Thr, p.Asp384Asn, p.Ala420Thr, and p.Ala747Thr) (Figure 4). Crucially, most of these mutations involve residues that are highly conserved among phylogenetically distant organisms (Ala420 presents valine or cysteine as alternative residues implying that this site requires conserved hydrophobicity) ( Figure 4B and Suppl. Figure S2). This underscores the functional importance of these sites, allowing us to suggest, as a common pathogenic consequence of the above non-conserved amino acid replacements, a defective binding of tRNA.
The p.Arg497His and p.Ala626Asp mutations fall near the binding pocket of the cognate valine. The first causes the replacement of an invariant arginine with a histidine that can also acquire a positive charge through protonation but presenting much lower pK a than the protonated arginine and thus the latter has stronger capability to maintain a positive charge also inside the protein environment.
In addition, the rigid aromatic ring of the histidine side chain does not reproduce the interactions of the flexible side chain of the native arginine. Although the site of the Ala626Asp mutation is apparently annuum, XM_016687220.1; P. x bretschneideri, XM_009342992.2) around the sites of the missense mutations discussed in the text (Thr367Ile, Ala379Thr, Asp384Asn, Ala420Thr, Arg497His, Ala626Asp, Ala747Thr). Residues that are invariant in this group of eukaryotes are shown in gray. C: homology model of VARS2. VARS2 protein (ribbons in different colors for the various functional regions), the residues affected by the missense mutations (yellow spheres), the bound cognate tRNA (tRNA-Val, light green ribbons and sticks), and the Val-AMP analogue (Val-AMS, magenta sticks) are shown. The pathogenic mechanism of these mutations can be inferred from their location: Thr367Ile, Ala379Thr, Asp384Asn, Ala420Thr, and Ala747Thr occur at protein sites relevant for the binding of the tRNA molecule, whereas the Arg497His and Ala626Asp mutations affect the binding pocket of the cognate valine not highly conserved, the alternative residues found in other organisms are either hydrophobic or serine or threonine: in contrast with aspartic acid, all these amino acids contain aliphatic portions that can preserve the local multiple interactions as engaged by Ala626 (Suppl. Figure S3). Furthermore, in the alignment ( Figure 4B and Suppl. Figure S2), this site never hosts a negatively charged residue like that introduced with the Ala626Asp replacement. This can be understood considering that such variant would cause salt-bridge shuffling, owing to its proximity to the ionic pair Arg274-Asp635. Based on these observations, Arg497His and Ala626Asp mutations are both expected to modify the conformation of the valine binding region and thus to impair enzyme function.

CONCLUSIONS
Here we describe the clinical and biochemical phenotype of 13 patients harboring bi-allelic pathogenic VARS2 variants. The common phenotype is characterized by a severe, early onset cardioencephalomyopathy associated with a combined OXPHOS defect in muscle. The brain MRI did not show any characteristic recognizable pattern across patients. Patient P1 has no evidence of cardiomyopathy; she and P5 (mild, later onset cardiomyopathy) are the only subjects who remain alive at ages 5 and 18 years, respectively. It is reasonable to infer from our cohort's data that a poor prognosis seems to correlate with the severity of cardiomyopathy/myopathy. P11 and P12, in which cardiomyopathy was not investigated, presented a severe encephalopathy with refractory epilepsy and died at 7 and 8 years, respectively. Indeed patient P5 also had a relatively mild neurological phenotype. These last observations are in line with the extreme clinical variability of mt-aaRS associated phenotypes, even though the described VARS2-associated phenotype looks rather homogenous. Structural and functional analyses clearly support the pathogenic role of the identified variants. The series of patients described here, reinforces previous findings and further delineates the range of clinical observations caused by mutations in VARS2, which should be investigated in early-onset mitochondrial encephalomyopathies or encephalocardiomyopathies.