ATP6V0C variants impair V-ATPase function causing a neurodevelopmental disorder often associated with epilepsy

Abstract

The vacuolar H⁺-ATPase is an enzymatic complex that functions in an ATP-dependent manner to pump protons across membranes and acidify organelles, thereby creating the proton/pH gradient required for membrane trafficking by several different types of transporters. We describe heterozygous point variants in ATP6V0C, encoding the c-subunit in the membrane bound integral domain of the vacuolar H⁺-ATPase, in 27 patients with neurodevelopmental abnormalities with or without epilepsy. Corpus callosum hypoplasia and cardiac abnormalities were also present in some patients. In silico modelling suggested that the patient variants interfere with the interactions between the ATP6V0C and ATP6V0A subunits during ATP hydrolysis. Consistent with decreased vacuolar H⁺-ATPase activity, functional analyses conducted in Saccharomyces cerevisiae revealed reduced LysoSensor fluorescence and reduced growth in media containing varying concentrations of CaCl₂. Knockdown of ATP6V0C in Drosophila resulted in increased duration of seizure-like behaviour, and the expression of selected patient variants in Caenorhabditis elegans led to reduced growth, motor dysfunction and reduced lifespan. In summary, this study establishes ATP6V0C as an important disease gene, describes the clinical features of the associated neurodevelopmental disorder and provides insight into disease mechanisms.

Mattison et al. identify ATP6V0C variants in patients with a neurodevelopmental syndrome. Using computational modelling and studies in Drosophila, yeast and worms, they show that the variants result in decreased V-ATPase function, providing evidence for their pathogenicity as well as insights into the underlying disease mechanisms.

Introduction

The vacuolar H⁺-ATPase (V-ATPase) is a highly conserved enzymatic complex that functions in an ATP-dependent manner to pump protons across membranes and acidify organelles. The V-ATPase is composed of a peripheral V₁ domain and an integral V₀ domain (Fig. 1). The V₁ domain is responsible for the hydrolysis of ATP creating the necessary energy to translocate protons through the V₀ domain via a rotational mechanism.¹ The V-ATPase plays a crucial role in many cellular processes involving membrane trafficking by creating a proton/pH gradient used by several different types of transporters.^1,2

Figure 1.

V-ATPase structure. The peripheral domain (V₁, uppercase letters, in grey) is the site of ATP binding and hydrolysis. The integral domain (V₀, lowercase letters, in purple, red, blue and yellow) transports protons across membranes. The c-ring (red) is composed of nine c-subunits (encoded by ATP6V0C) and one c′′-subunit (encoded by ATP6V0B, not shown) and rotates after ATP hydrolysis to bring protons to ATP6V0A (blue). ATP6V0A possess two hemi-channels and a buried arginine residue (p.R735) that are required along with p.E139 in ATP6V0C for proton translocation.¹

The human V-ATPase, comprising 13 different subunits, is encoded by 23 genes (Supplementary Table 1). This genetic redundancy allows for the formation of tissue-specific V-ATPase complexes, including in synaptic vesicles (SVs) where it creates the necessary proton/pH gradient to load various neurotransmitters.^1,2 To date, 12 genes corresponding to eight different subunits of the V-ATPase have been associated with human disease (Supplementary Table 1).^3–13 Early onset epilepsy has been observed in patients with variants in either ATP6V1A or ATP6V0A1, with heterozygous, de novo variants leading to less severe presentations when compared to patients with biallelic variants.^14–16 Pathogenic variants in ATP6V1B2 can cause epileptic conditions such as Zimmerman–Laband (MIM#616455) and DOORS (deafness, onychodystrophy, osteodystrophy, intellectual disability and sometimes seizures) syndromes, or deafness and nail dysplasia without epilepsy (DDOD, MIM#124480).^11,12,17,18 An accessory protein to the V-ATPase, encoded by ATP6AP2, is associated with X-linked syndromic intellectual disability that can present with or without epilepsy (MIM#300423).^19,20

ATP6V0C, a three-exon gene (Fig. 2A) located on human chromosome 16p13.3, encodes the 155 amino acid c-subunit of the V₀ domain which along with the c′′ subunit (encoded by ATP6V0B) forms the intramembrane c-ring that facilitates the movement of protons across the membrane (Fig. 1).²¹ The process of proton translocation is reliant on a glutamate residue at position 139 (p.E139) in ATP6V0C as well as an arginine residue (p.R735) in ATP6V0A.¹

Figure 2.

Location and conservation of ATP6V0C variants. (A) Exon/intron structure of ATP6V0C. Boxes represent exons with black denoting coding regions. Scale bar = 100 bp length. (B) Protein alignment showing conservation of affected residues (highlighted in yellow). Glutamate residue (p.E139) required for proton transport is in bold. The following protein sequences were used in the alignments: H. sapiens, NP_001685.1; M. musculus, NP_001348461.1; D. rerio, NP_991117.7; D. melanogaster, NP_476801.1; C. elegans, NP_499166.1; S. cerevisiae, NP_010887.3. (C) Lollipop plot showing the transmembrane structure (green) and location of variants throughout ATP6V0C. Patient missense variants are indicated above in red. Missense (blue) and synonymous (grey) variants observed in gnomAD are shown below. Based on UniProt accession P27449. There is a significant enrichment of patient variants in TM4 (P = 0.006, Fisher’s exact test). (D) Plot showing tolerance of missense variants across ATP6V0C. The missense tolerance ratio (MTR) was calculated using 21 codon window sizes. A MTR score of <1 indicates intolerance to missense variation. Dashed lines on the plot denote ATP6V0C-specific MTRs: green = 5th percentile, yellow = 25th percentile and black = 50th percentile.

We previously described patients with developmental delay, intellectual disability, microcephaly and seizures with 16p13.3 microdeletions encompassing a minimal overlapping region that included TBC1D24, ATP6V0C and PDPK1.²² By reviewing the known function(s) and expression patterns of genes in the minimal overlapping region, we proposed haploinsufficiency of ATP6V0C as the primary contributor to the clinical features of 16p13.3 microdeletion syndrome.²³ However, we did not provide any functional evidence to support our claim.

Most recently, Ittiwut et al.²⁴ reported a de novo stop-loss variant in ATP6V0C in an individual with epilepsy and intellectual disability. Analysis of RNA derived from the patient’s leucocytes revealed that, as expected, the mutant transcript escaped nonsense mediated decay (NMD). The authors proposed haploinsufficiency as the probable pathomechanism given the observed decrease in mRNA levels, however, a dominant negative effect is also possible given the transcript escape from NMD. Hence, thus far it is unknown whether ATP6V0C missense variants are a cause of human disease, and the mechanistic basis of ATP6V0C-associated human disease is unclear.

In this study, we report the identification of heterozygous ATP6V0C missense variants in 27 patients with a novel syndrome of developmental delay, epilepsy and intellectual disability. We present multiple lines of computational and functional analyses to demonstrate that these variants are pathogenic and disrupt V-ATPase activity.

Materials and methods

Identification of individuals with ATP6V0C variants

Authorization was obtained at each site to release deidentified patient medical information to study investigators and when applicable informed consent was obtained through protocols approved by the institutional review boards at each site of patient recruitment. Patients with ATP6V0C variants were identified via GeneMatcher, and by interrogating the 100 000 Genomes database, the Deciphering Developmental Disorders (DDD) study and ClinVar.^25–27 We also screened whole-exome sequencing data from epilepsy patients referred for genetic testing at EGL Genetics. Three patients (Patients 2, 13 and 27) were previously reported in other publications.^24,25,28 For Patients 2 and 27, clinical information was obtained from the previous publications, whereas the referring clinician provided clinical information for Patient 13. Patient 5 was reported in ClinVar and clinical information was provided by the depositing organization (Supplementary Table 2). Clinical information for all other patients was collected using a custom form provided to each site.

Sequencing and analysis of sequence data

Whole-exome or -genome sequencing was performed on patient DNA extracted through standard protocols. All libraries were sequenced on Illumina HiSeq systems. Sequence alignment and variant calling were performed at each site and further details are provided in Supplementary Table 2. When possible, ATP6V0C variant segregation was confirmed with Sanger sequencing using standard protocols.

Lollipop and Missense tolerance ratio diagram

The Lollipop diagram was created as previously described using the UniProt accession number, P27449.^29,30 Non-synonymous and synonymous population variants were downloaded from gnomAD (v.2.1.1).³¹ Resulting diagrams of gnomAD and patient variants were merged into a single image for ease of visualization, and the locations of the transmembrane domains were superimposed over the resulting image. A missense tolerance ratio (MTR) plot was generated using MTR-Viewer v.2, with a codon window size of 21, on the ENST00000330398 transcript.³²

Drosophila studies

The Drosophila orthologue, Dmel\Vha16–3 (CG32090), was knocked-down using pan-neuronal (elaV-GAL4) expression of a gene-specific RNAi construct (VDRC-102067), provided by the Vienna Drosophila Resource Center. As a control, an RNAi to GFP was used. Seizure-like behaviour was elicited using electroshock of wall-climbing third instar larvae as described in Marley and Baines’ study.³³ Drugs, solubilized in dimethyl sulphoxide, were fed to larvae by mixing (at 2 mM) in molten fly food, which was then allowed to cool and set before being seeded with first instar larvae.

In silico variant modelling

Patient and gnomAD variants were displayed in the context of a structure for the V₀ domain of human V-ATPase [Protein Data Bank (PDB) ID 6wlw].^21,34 PyMOL and Swiss PDB Viewer were used for visualization of protein structure.³⁵

Saccharomyces cerevisiae strains and plasmids

E. coli and yeast manipulations were performed following standard molecular biology protocols.³⁶ The vma3::kanMX yeast strain (vma3Δ, catalogue no. YSC6273–201929081) was obtained from GE Dharmacon and is isogenic with BY4741 (MATa his3Δ leu2Δ met15Δ ura3Δ). Plasmids are listed in Supplementary Table 3. A plasmid (pKM16) containing the promoter and wild-type open reading frame for VMA3 (the yeast orthologue of ATP6V0C) was generated by amplifying a 924 bp fragment from S. cerevisiae gDNA and cloning it into pRS316 using BamHI and SacI (Forward cloning primer: 5′; taagcaggatccagcaatgaaataggccgtctac, Reverse cloning primer: 5′; taagcagagctccttgaaatgaggtagtttgg).³⁷ pKM16 was used as the backbone to generate all variants via conventional cloning techniques or the QuikChange XL Site-Directed Mutagenesis Kit (Agilent Technologies). Sanger sequencing was performed to confirm the presence of each variant as well as the absence of unwanted substitutions.

Plasmids were transformed into the vma3Δ strain and selected on plates containing synthetic minimal media plus dextrose supplemented to select for a URA3 plasmid (SD-ura). Selected transformants were maintained on SD-ura throughout the course of experiments.

Serial dilution spotting assay

Liquid cultures of transformants were grown at 30°C overnight in SD-ura adjusted to a pH of 5.5. Then 1 optical density (OD₆₀₀) per ml of cells was collected and suspended in dH₂0. Serial 10-fold dilutions were spotted onto SD-ura plates. Plates were imaged after incubation at 30°C for 48 h.

LysoSensor uptake and confocal imaging

Cells were collected at OD₆₀₀ 0.6–1.0 and incubated with LysoSensor Green DND-189 (Invitrogen, catalogue no. L7535) as previously described.³⁸ Cells were resuspended in 1× PBS (pH 7.6) to an OD₆₀₀ of 0.6, deposited on 1.5% agarose pads, and visualized immediately. Images were taken at room temperature using a confocal laser scanning microscope (A1R HD25, Nikon) with an Apo TIRF ×60 1.49 numerical aperture (NA) oil immersion lens, WD 0.12 mm (Nikon). Images were acquired using NIS-Elements (AR 5.21.02, Nikon) and processed using FIJI. All cells visible in the DIC channel were selected, the selection was copied to the FITC channel and the mean grey area was calculated for each cell. Measurements were corrected for background signal by subtracting the mean grey area of a background only selection from each image. Two biological replicates, in separate experiments, totalling 71–132 cells for each variant were analysed.

Generation of growth curves

Liquid cultures of transformants were grown at 30°C overnight in SD-ura adjusted to a pH of 5.5. Overnight cultures were diluted 1:20 in dH₂0 and further diluted 1:10 in YPD, pH 5.5, with 5, 100 or 200 mM CaCl₂. OD₆₀₀ measurements were taken every 15 min for 30 h at 30°C with shaking using an ELx808 plate reader (BioTek). Three independent transformants were assayed in triplicate for each variant. The R package, GrowthCurver, was used to calculate the empirical area under the curve (eAUC) for each replicate.³⁹

Caenorhabditis elegans studies

All C. elegans strains were cultured and handled as per standard methods. All experiments were carried out at 20°C. Mutations were knocked-in via CRISPR–Cas9, and homozygous worms were studied. Strains used in this project are described in Supplementary Table 4.

Paralysis assays

In three separate experiments, 30–40 L4 worms (in triplicate) were picked to standard NGM plates, either with 51 mM NaCl (physiological conditions) or with 200 or 300 mM NaCl, and scored daily for paralysis starting the day after they were picked. A total of 270–360 worms were tested per condition. Worms were counted as paralysed if they failed to move their body on prodding with a platinum wire, and were considered dead if they failed to move their head and showed no pharyngeal pumping when prodded. Dead or lost animals were censored from statistical analyses. Worms were transferred every 2 days to avoid progeny.

Lifespan assays

Lifespan experiments were conducted similarly to paralysis assays. Two separate experiments were performed using 35–40 worms in triplicate, leading to a total 315–360 worms tested per condition. Worms were counted every second day from Day 1 of adulthood until death. Lost animals were censored from statistical analyses, paralysed worms were not censored and were kept until death.

Liquid culture motility assays

Synchronized Day 1 adult worms were transferred into 200 µl of M9 buffer with or without 350 mM NaCl in each well of a 96-well plate, to a density of 30 worms per well. Motility was automatically analysed for 4 h using a WMicrotracker-One plate reader (PhylumTech).

WormLab analysis

Synchronized Day 1 adult worms were video recorded for 30–33 s using a Leica Stereomicroscope S9i. Automated movement and worm size analyses were conducted using WormLab software (MBF Biosciences), which tracks individual worms from the recorded videos. Activity index is defined by the brush stroke (area ‘painted’ by the animal's body in a single complete stroke) normalized by the time taken to perform two strokes. Wave initiation rate is defined as the number of body waves initiated from either the head or the tail per minute. Swimming speed was measured over a two-stroke interval.⁴⁰ For body size analysis, worms were recorded on bacteria free NGM plates. For swimming parameters, worms were placed in M9 with 500 mM NaCl and recorded 30 min later. At least 50 worms were recorded in two independent experiments.

Aldicarb sensitivity assays

Worms were grown on standard NGM plates until Day 1 of adulthood and then transferred to NGM plates containing 1 mM aldicarb. Paralysis was assayed every 30 min for 2 h. Animals were counted as paralysed if they failed to move on prodding with a platinum wire.

Statistical analyses

Drosophila recovery time was analysed using one-way ANOVA with Dunnett’s post hoc test for multiple comparisons. Recovery time after drug treatment was normalized to the vehicle only control. A Fisher’s exact test was used to demonstrate the presence of a variant hot-spot in the fourth transmembrane domain of ATP6V0C. For LysoSensor, fluorescence values for each cell measured were normalized to the mean fluorescence of the wild-type rescue. For the growth rate assay, eAUC values were normalized to the mean eAUC of the wild-type rescue within each plate. A one-sample t-test was used to compare the mean fluorescence and eAUC for each variant to a hypothetical mean of 100 (representing the wild-type rescue). Significance levels were corrected for multiple comparisons using a Bonferroni correction (LysoSensor and growth at 5 mM, α = 0.0003125; 100 mM, α = 0.00714; 200 mM, α = 0.00833). A two sample t-test (two-tailed) was used to compare p.G103S to p.F137L at 200 mM. Paralysis curves and lifespan assays were compared using the log-rank (Mantel–Cox) test. The liquid culture motility assays (WormTracker) results were analysed using two-way ANOVA to compare each variant to the wild-type N2 strain. WormLab results were analysed using a one-way ANOVA with a Dunnett’s post hoc test for multiple comparisons to compare each variant to the wild-type N2 strain. The WormLab data is presented as box and whisker plots indicating minimal and maximal data points. Normalization and statistical analyses were carried out using Prism v.9.0 (GraphPad Software Inc.). All data is presented as mean ± SEM and α = 0.05 was used unless otherwise noted.

Data and reagent availability

All strains and plasmids (Supplementary Tables 3 and 4) used in this study are available on request. The human datasets presented in this article are not publicly available due to ethical and privacy restrictions. Requests to access the human datasets should be directed to the corresponding authors.

Results

Identification of ATP6V0C variants in patients

We identified 27 patients with heterozygous ATP6V0C variants through GeneMatcher, 100 000 Genomes database, the DDD study, ClinVar, published literature and EGL Genetics. Of the 27 patients, 22 had missense substitutions (18 variants were unique), four had frameshifting variants and one had a stop-loss variant (Table 1). The variants p.A138P, p.A149T and p.L150F were recurrent, each being seen in two or more unrelated individuals. Different substitutions at p.A95 and p.L150 were also observed. Patients 4, 9, 15 and 19 were each found to be mosaic for their identified ATP6V0C variant.

Table 1.

Clinical presentation of patients with ATP6V0C variants

Patient	Varianta	CADD Scoreb	Inheritance	Seizures (age at onset)	Seizure types	Developmental delay	Intellectual disabilityc
1d	c.85G > A; p.G29S	26.2	de novo	NA	NA	NA	NA
2e	c.134_135delCT; p.(S45CfsTer37)	NA	de novo	Yes (7 mo)	GTCS, At, FDS, Myo, T	NA	Severe, with regression
3	c.143G > C; p.R48P	27.2	de novo	Yes (18 mo)	NA	Yes, motor and speech	NA
4	c.158T > G; p.M53R	25.9	de novo, mosaic	No	NA	Yes	Too young for evaluation
5f	c.172C > G; p.P58A	23.1	NA	Yes	Cryptogenic focal	Yes, psychomotor	NA
6	c.188G > C; p.G63A	23.9	de novo	Yes (8 mo)	Infantile spasms, GTCS, At, Myo	Yes, non-verbal	Severe
7	c.214delG; p.(V72WfsTer9)	NA	de novo	Yes	Infantile spasms	Yes	Too young for evaluation
8	c.220G > T; p.V74F	26	de novo	Yes (12 mo)	GTCS, Ab, FOA	Yes, motor and speech	Severe
9	c.283G > A; p.A95T	26.2	de novo, mosaic	Yes (10 mo)	GTCS, FOA, Febrile	Yes, regression to non-verbal	Profound
10	c.283G > C; p.A95P	24.6	de novo	Yes (10 mo)	GTCS, staring	Yes, motor and non-verbal	Too young for evaluation
11	c.284C > T; p.A95V	23.7	de novo	Yes (5 mo)	Febrile, Ab, Myo, T (nocturnal)	No	Moderate
12d	c.294C > A; p.S98R	13.2	de novo	NA	NA	NA	NA
13g	c.340_355del16; p.(D115AfsTer12)	NA	de novo	Yes (16 mo)	Focal with secondary generalization	Yes, speaks only in short sentences	Yes
14	c.352_362delins; p.(V118HfsTer19)	NA	de novo	Yes (30 mo)	GTCS, focal to bilateral TCS, Ab, FIA	Yes, motor and speech	Mild, regression in adulthood
15	c.395T > A; p.I132N	25.3	de novo, mosaic	Yes (12 yr)	NA	Yes	Yes
16	c.409T > C; p.F137L	25.2	de novo	Yes (13 mo)	GTCS, Myo, At, FOA	Yes, motor and non-verbal	Profound
17	c.412G > C; p.A138P	25.5	de novo	NA	TCS	Yes	NA
18	c.412G > C; p.A138P	25.5	de novo	Yes (6 mo)	GTCS, multifocal	Yes, motor and speech	Too young for evaluation
19	c.425G > A; p.G142D	24.7	de novo, mosaic	No	NA	Yes, motor and non-verbal	Too young for evaluation
20	c.440T > C; p.I147T	23.3	de novo	Yes	NA	Yes, speech	Profound
21	c.445G > A; p.A149T	24.3	de novo	Yes (38 mo)	Febrile, TCS, Myo, Ab	Yes, motor and speech	Mild
22	c.445G > A; p.A149T	24.3	de novo	Yes (6 mo)	GTCS, At, TCS	Yes, motor and speech	Mild
23d	c.448C > T; p.L150F	25.1	NA	NA	NA	NA	NA
24	c.448C > T; p.L150F	25.1	de novo	Yes (6 mo)	Infantile flexor spasms, T (w/asymmetrical limb stiffening)	Yes, motor and speech	Profound
25	c.448C > T; p.L150F	25.1	de novo	Yes (6 yr)	GTCS	Yes, motor and speech	Severe
26	c.449T > C; p.L150P	24.7	de novo	Yes (18 mo)	Febrile, GTCS, T, At, Ab	Yes, fine motor	Mild
27h	c.467A > T; p.(Ter156LeuextTer35)	NA	de novo	Yes (24 mo)	GTCS, TCS, At, afebrile	No	Yes, with regression

Ab = absence; At = atonic; FDS = focal dyscognitive seizures; FOA = focal onset aware (partial); FIA = focal impaired aware; GTCS = generalized tonic-clonic seizures; mo = months; Myo = myoclonic; NA = not available; T = tonic; TCS = tonic-clonic seizures.

Based on reference sequence NM_001694.4.

Scores obtained using CADD GRCh37-v1.6.

Intellectual disability can usually be first assessed at 5 years of age.

Patients 1, 12 and 23 have severe neurodeveopmental diseases but detailed clinical information was unavailable.

Previously published as Patient T1911 in Carvill et al.²⁸

ClinVar variant: VCV000870676.

Previously published as DDD4K.04123 in DDD study.²⁵

Previously published in Ittiwut et al.²⁴

Multiple lines of evidence support the pathogenicity of the identified patient variants. First, all patient variants were absent from the Genome Aggregation Database (gnomAD, v.2.1.1), which is compilation of exome and genome sequencing data from large-scale sequencing projects for which efforts have been made to remove individuals affected by severe paediatric disease. Second, in 24 patients where biologic parent DNA was available, the variants were found to have occurred de novo. Third, all missense substitutions affected highly conserved residues (Fig. 2B), with 17 out of the 18 unique missense variants having CADD scores that placed them in the top 1% of predicted deleterious variants. Last, ATP6V0C has a predicted intolerance to missense and loss of function variation with only 21 population missense variants observed in gnomAD compared to the expectation of 108 (observed/expected = 0.19) and 0 loss of function variants observed compared to the expectation of 4.5 (Fig. 2C and D). In addition, all ATP6V0C gnomAD variants are observed at low frequencies (three times or less).⁴¹ Taken together, when the ACMG/AMP variant classification guidelines are applied, all ATP6V0C variants identified in the patients are predicted to be likely pathogenic or pathogenic (Supplementary Table 2).⁴²

ATP6V0C variants cause a human syndrome of developmental delay, epilepsy and intellectual disability

The primary clinical presentation of the identified patients was developmental delay with early onset epilepsy and intellectual disability (Table 1). The mean age of seizure onset was 24.6 ± 8.0 months, with 14 of 18 patients for whom this information was available having onset before 24 months. On the basis of clinical information from 19 patients, the most common seizure types observed were generalized tonic-clonic (12/19), focal (7/19), atonic (6/19) and myoclonic (5/19). Intellectual disability, ranging from mild to severe, was seen in 16/16 patients who were old enough for a formal diagnosis and for whom this information was available. Development delay was seen in 21/23 patients. Twenty-one patients had MRIs with 13 showing abnormalities (Supplementary Table 2). Common findings in MRIs included agenesis/hypoplasia of the corpus callosum (Supplementary Fig. 2) or cerebellar vermis (Patients 5, 6, 18, 20 and 26), and delayed myelination (Patients 6, 18, 22 and 24). Four patients were reported to have cardiac abnormalities: Patient 3 had pulmonary valve stenosis, Patient 6 had a thickened left ventricular wall, Patient 7 had a heart murmur and Patient 13 exhibited several cardiac defects including hypertrophic cardiomyopathy, mitral valve prolapse and mild to moderate mitral valve regurgitation. Patients 5 and 24 showed dental enamel defects, with Patient 24 lacking dental enamel. It should be noted that Patient 7 also has a de novo 2.3 Mb deletion in 20q11.22–11.23, which has been associated with developmental delay and dysmorphism, and Patient 13 has biallelic variants in LZTR1, which has been associated with Noonan-like syndrome (MIM#605375).⁴³

Collectively, these data show that ATP6V0C variants cause a human syndrome of developmental delay, intellectual disability and epilepsy. Furthermore, as most individuals were ascertained on the basis of genotype (i.e. having a variant in ATP6V0C), their phenotypic convergence on reverse phenotyping further supports the pathogenicity of the variants described in this study.⁴⁴

ATP6V0C knockdown in Drosophila results in seizure-like behaviour

On the basis of our hypothesis that haploinsufficiency of ATP6V0C drives the neurological phenotype of 16p13.3 microdeletion syndrome and the identification of patients with frameshift variants, we first tested the consequences of ATP6V0C knockdown in Drosophila.²³ The orthologous protein in Drosophila, Dmel\Vha16–3 (CG32090), shows 78% amino acid identity to ATP6V0C. CG32090 was knocked-down via pan-neuronal expression of a gene-specific RNAi construct (VDRC-102067). The same pan-neuronal driver line (elaV-GAL4) driving expression of GFP RNAi (elaV > GFP RNAi), and the homozygous 102067 RNAi line without the elaV-GAL4 driver (102067 RNAi) were used as controls. Following knockdown of CG32090 (elaV > 102067 RNAi), wall-climbing third instar larvae showed a significant increase in recovery time (i.e. longer seizure-like behaviour) after electroshock (P < 0.0001, one-way ANOVA; Fig. 3A). Pretreatment of these larvae with a variety of established antiepileptic drugs resulted in significant reductions in recovery time with levetiracetam and topiramate (P < 0.001), and to a lesser extent with lamotrigine and valproate (P < 0.01 and <0.05, respectively; Fig. 3B). Phenytoin, at the same concentration (2 mM in fly food) did not significantly alter recovery time. These results are consistent with the hypothesis that haploinsufficiency of ATP6V0C contributes to seizures.

Figure 3.

Knockdown of the Drosophila orthologue of ATP6V0C increases seizure duration. (A) Pan-neuronal (elaV-GAL4) RNAi-mediated knockdown of Dmel\Vha16–3 (CG32090) using RNAi (elaV > 102067 RNAi) is sufficient to increase the recovery time (RT) of third instar larvae to electroshock-induced seizure. Controls expressed GFP RNAi via elaV-GAL4 (elaV > GFP RNAi) or the RNAi (102067) without a driver (102067 RNAi). (B) Seizure induction due to expression of 102067 RNAi is preferentially rescued by pretreatment of larvae with levetiracetam (LEV) or topiramate (TOP). Lamotrigine (LAM) and valproate (VAL) were also effective, but not phenytoin (PHY). RT was normalized to a vehicle (dimethyl sulphoxide) only control. Data shown as mean ± SEM. One-way ANOVA with post hoc comparison (Dunnett’s); *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

ATP6V0C variants are predicted to interfere with V-ATPase rotary mechanism

Higher conservation across ATP6V0C orthologues was seen at sites of patient variants compared to gnomAD variants (Supplementary Fig. 1). To understand the basis of pathogenicity of the ATP6V0C patient missense variants, we first turned to in silico modelling. On hydrolysis of ATP, the c-ring (comprising nine copies of ATP6V0C and one copy of ATP6V0B) rotates within the membrane delivering protons to the ATP6V0A subunit (encoded by ATP6V0A) for transport across the membrane (Fig. 1).²¹ Transmembrane (TM) domains 2 and 4 of ATP6V0C are outward facing and interact with ATP6V0A during this rotational mechanism.^45,46 The location of patient variants shows an enrichment within TM domains and the presence of a ‘hot-spot’ in the fourth TM of ATP6V0C (P = 0.006, Fisher’s exact test; Fig. 2C). When viewed structurally, some patient and gnomAD variants are located at sites of packing between c-ring subunits; however, many more patient variants are located outward facing from the c-ring so as to potentially interfere with interactions between mutant ATP6V0C subunits and the ATP6V0A subunit (Fig. 4A and B). These data raise the possibility that outward-facing ATP6V0C missense variants may have a dominant negative effect.

Figure 4.

Molecular modelling of patient and gnomAD variants. (A) Structure of part of the V₀ region of human V-ATPase (PDB 6wlw).²¹ Sites of patient (purple) and gnomAD (green) variants are shown superposed on ribbon backbone for two ATP6V0C subunits (gold), one next to the ATP6V0A subunit (cyan) and one on the opposite side of the c-ring. The back part of the c-ring is filled with grey, and the front part has been omitted for clarity. (B) Isolated view of the interaction between ATP6V0C variants and ATP6V0A. The functional amino acid p.E139 is also displayed (pink).

ATP6V0C patient variants are deleterious in yeast

Budding yeast, S. cerevisiae, possesses an ATP6V0C orthologue, VMA3, which shares 72% amino acid identity and a conserved four transmembrane protein structure. S. cerevisiae has been previously used to study the functional effects of variants in other V-ATPase subunits.^8,47–50 Given that all identified patient missense variants affected residues that are conserved between human and yeast, we expressed 12 of the patient variants in VMA3 using a centromeric plasmid in a vma3Δ yeast strain (Supplementary Fig. 1 and Supplementary Table 3). Six additional variants were identified after completion of these experiments and, therefore, were not modelled in yeast.

We also examined the functional effects of three population variants in ATP6V0C from gnomAD (p.R48W, p.G103S, p.M131I; Supplementary Fig. 1 and Supplementary Table 5) on V-ATPase function. p.R48W was chosen as it effects the same residue as the p.R48P patient variant. p.G103S and p.M131I were chosen as they have the highest CADD scores for variants seen twice and once, respectively, in gnomAD. The altered residues are also conserved between human and yeast (Supplementary Fig. 1). All ATP6V0C variants in gnomAD (21 total) are rare (Supplementary Fig. 1), being seen no more than three times out of ∼250 000 alleles. In addition, we also generated and tested p.E139A, which removes the glutamate residue necessary for V-ATPase function.¹

To first establish the ability of the vma3Δ strain to grow when transformed with plasmids containing patient or population variants, a serial dilution spot assay was performed on SD-ura plates. We confirmed the ability of all transformants to grow under no selective V-ATPase pressure, thereby allowing us to examine V-ATPase function in our yeast model (Supplementary Fig. 3).

LysoSensor Green DND-189 (pK_a ∼5.2) is an acidotropic dye that accumulates in the membranes of vacuoles. On protonation, quenching of the fluorescent probe is relieved and fluorescent intensity increases in a pH dependent manner. It has been previously demonstrated that V-ATPase activity in yeast and fluorescent intensity of LysoSensor Green are correlated.³⁸ We looked for rescue of V-ATPase function by transformation of each patient or population variant into the vma3Δ strain. Nine patient variants and p.E139A resulted in little to no fluorescence (Fig. 5A and Supplementary Fig. 4). Two patient variants (p.G63A and p.L150F) and two gnomAD variants (p.R48W and p.G103S) showed intermediate levels of fluorescence intensity compared to the wild-type rescue. One patient variant (p.F137L) showed levels of fluorescence intensity that were comparable to wild-type rescue. Overall, 11 of 12 patient variants and all gnomAD variants elicited significant decreases in fluorescence intensity when compared to the wild-type rescue (P < 0.01, one-sample t-test; Fig. 5A).

Figure 5.

Patient variants show reduced V-ATPase function. (A) Quantification of average fluorescent intensity for each variant in the LysoSensor assay. Variants are grouped based on their location within or proximity to the nearest transmembrane (TM) domain. Data were normalized with mean of wild-type as 100% (denoted by dotted line) and mean of empty vector as 0%. Data shown as mean ± SEM (n = 71–132 cells per variant). Box and whisker plot of these data is presented in Supplementary Fig. 4. (B–E) Growth curves of vma3Δ S. cerevisiae expressing patient or gnomAD variants when grown in YPD, pH 5.5 with 5 mM CaCl₂. In all panels, wild-type is shown in black and the empty vector in grey. Mean of nine replicates per construct is shown with error bars omitted for clarity. Variants are grouped on the basis of their location within or proximity to the nearest TM domain. (F) eAUC was calculated using Growthcurver.³⁹ Data were normalized within each plate with wild-type as 100% (denoted by dotted line) and empty vector as 0% and is shown as mean ± SEM. A one-sample t-test to a hypothetical mean of 100 was conducted with a Bonferroni correction (adjusted α level = 0.0003125). **P < 0.01, ***P < 0.001, ^#P < 0.0001.

V-ATPase, and therefore ATP6V0C, function is required for yeast to grow at increased CaCl₂ concentrations.^47,51 To further examine the consequences of patient ATP6V0C variants on V-ATPase function, transformants were used to inoculate YPD with 5 mM CaCl₂ and growth curves were generated for each variant (Fig. 5B–E). Eight patient variants and p.E139A showed negligible growth at 5 mM CaCl₂. The remaining four patient variants and all gnomAD variants showed varying degrees of growth at 5 mM CaCl₂. Overall, significantly less growth compared to wild-type rescue was seen for 11 of 12 patients and the three gnomAD variants tested (P < 0.0001, one-sample t-test; Fig. 5F). The results seen at 5 mM CaCl₂ mirrored those seen with LysoSensor uptake (r² = 0.7921; Supplementary Fig. 5).

Next, patient variants that grew at 5 mM CaCl₂ along with the three gnomAD variants were tested at 100 and 200 mM CaCl₂ to determine whether a higher concentration of calcium would provide further separation of variants relative to the wild-type rescue. At 100 mM CaCl₂, p.L150F showed almost no growth compared to wild-type (P < 0.0001, one-sample t-test; Supplementary Fig. 6A and B) while the three other patient variants (p.G63A, p.V74F, p.F137L) showed intermediate growth relative to the wild-type rescue. The growth of one gnomAD variant (p.M131I) was similar to the wild-type rescue, while p.R48W and p.G103S both showed less growth relative to the wild-type rescue. The three patient variants with intermediate growth at 100 mM CaCl₂ and the three gnomAD variants were then tested at 200 mM CaCl₂ (Supplementary Fig. 6C and D). Significantly less growth was seen with the three patient variants compared to the wild-type rescue (P < 0.0001), while two gnomAD variants (p.R48W and p.M131I) were comparable to the wild-type rescue. p.G103S showed less growth relative to the other gnomAD variants (p.R48W and p.M131I), but still yielded a significantly larger eAUC (Supplementary Fig. 6D) compared to the best growing patient variant at 200 mM CaCl₂, p.F137L (59.22 ± 2.278 versus 45.63 ± 2.243, P = 0.0006, two sample t-test).

Assessment of three patient variants in C. elegans

Next, we assessed the effects of a subset of patient ATP6V0C variants on neurological function using C. elegans. Worms express three orthologous genes to ATP6V0C in neurons, vha-1, vha-2 and vha-3. VHA-2 and VHA-3 have identical amino acid sequences and share 66.7% amino acid identify with ATP6V0C, while VHA-1 has slightly less homology to ATP6V0C at 63% amino acid identity. The variants that were selected for further analysis were distributed throughout the protein, showed tolerance to 5 mM CaCl₂ in the yeast growth assay and were identified in patients with severe neurocognitive deficits and poly-medicated epilepsy. Specifically, we studied p.F137L (corresponding to p.F143L in VHA-2) and p.G63A and p.L150F (corresponding to p.G69A and p.L156F, respectively, in VHA-3). A fourth strain carrying the p.A95T variant (corresponding to p.A101T in VHA-2) was generated but caused sterility in homozygous worms and was not studied further.

Worms expressing each variant were shorter and smaller than N2 (wild-type) controls at Day 1 of adulthood, indicating a morphological delay even in ideal physiological conditions (P < 0.05, one-way ANOVA; Fig. 6A and B). When tested in liquid physiological M9 over a 4-h period, movement of Day 1 young adult worms with each mutation was comparable to N2 worms (Fig. 6C). However, when motor function was examined under osmotic stress conditions (350 mM NaCl), mutants expressing p.G63A and p.F137L exhibited significantly reduced movement scores when compared to N2 worms (P < 0.01, two-way ANOVA; Fig. 6D). Although mutants expressing p.L150F also exhibited less movement than N2 worms, this difference was not statistically significant (P = 0.0869).

Figure 6.

Expression of patient variants in C. elegans exacerbate motor dysfunction and reduce lifespan. (A and B) WormLab analysis of body length and size at Day 1 of adulthood. All mutants are shorter and smaller than N2 controls. (C and D) Automated analysis of worm movement in liquid culture by WormTracker software. (C) In physiological M9 solution, all mutants show no motor deficits. (D) In presence of 350 mM NaCl concentration the p.G63A (P < 0.0001) and p.F137L (P < 0.0062) mutants show reduced movement scores in liquid culture over 270 min. Reduced movement was also observed with the p.L150F variant, but this difference was not statistically significant (P = 0.0869). (E) All mutant strains showed increased paralysis over 14 days compared to N2 controls (n = 313–317/strain, P < 0.0001). (F and G) In presence of osmotic stress (200 or 300 mM NaCl) the paralysis phenotype is exacerbated, leading to almost 100% paralysis after 8 days for the p.G63A strain (n = 246–260/strain, P < 0.0001). (H) All mutant strains exhibited reduced lifespan compared to N2 controls (n = 219–233/strain, P < 0.0001). (I and J) All mutant strains have reduced lifespans in presence of osmotic stress compared to N2 controls (200 mM NaCl: n = 182–228/strain. 300 mM, P < 0.0001) (300 mM NaCl: n = 200–244/strain, P < 0.0001). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 compared to N2 controls.

We next compared paralysis and lifespan of each mutant strain with N2 worms when maintained on NGM plates under physiological conditions or exposure to osmotic stress (200 and 300 mM NaCl). All mutants showed greater levels of paralysis when compared to N2 worms over a 14-day period, and these differences were strongly exacerbated in the presence of osmotic stress [P < 0.0001, log-rank (Mantel–Cox) test; Fig. 6E–G]. Lifespans of the mutant strains were also reduced when compared to N2 worms when maintained under physiological conditions and osmotic stress [P < 0.0001, log-rank (Mantel–Cox) test; Fig. 6H–J].

To compare fine motor phenotypes, we exposed the mutant strains and N2 worms to liquid M9 with 500 mM NaCl for 30 min and analysed movement using WormLab. Activity index and wave initiation were significantly increased in the mutant strains when compared to N2 worms (P < 0.05), but swimming speeds were not significantly altered, suggesting increased but uncoordinated movements (P > 0.05, one-way ANOVA; Fig. 7A–C).

Figure 7.

Patient mutations cause increased uncoordinated movement and neuronal signalling dysfunction in C. elegans. (A–C) Analysis of fine motor movement of worms after 30 min in 500 mM NaCl liquid culture. Mutants show increased activity index and wave initiation (A and B), but swimming speed was not significantly altered (C). (D) Synaptic transmission was evaluated by exposing Day 1 adult worms to aldicarb. Worms were scored over a 2-h period for paralysis. All mutants were hypersensitive to aldicarb treatment compared to N2 worms (n = 236–296/strain). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 compared to N2 controls.

To test whether the mutant strains have an impairment in nervous system signalling, we added an acetylcholinesterase inhibitor, aldicarb, to NGM plates and scored the number of paralysed worms over a 2-h period. Aldicarb causes an accumulation of acetylcholine in neuromuscular junctions resulting in muscle hypercontraction and acute paralysis, and can be used to evaluate whether there is dysfunction of either GABA or acetylcholine signalling.^52,53 To confirm proper aldicarb effect, we included unc-47(e307) and unc-64(e246) mutants. Unc-47(e307) mutants are hypersensitive to aldicarb due to the lack of a vesicular GABA transporter gene (orthologous to SLC32A1 in humans) required for GABA transmission.⁵⁴ Unc-64(e246) (orthologous to STX1A in humans) mutants have reduced cholinergic neurotransmission, making them resistant to aldicarb-induced paralysis.^55,56 Worms expressing each patient variant showed greater paralysis in presence of aldicarb, compared to N2 worms [P < 0.0001, log-rank (Mantel–Cox) test; Fig. 7D].

Discussion

In this study, we report the identification of heterozygous ATP6V0C variants in 27 patients with neurodevelopmental phenotypes. In general, this cohort of patients presented with development delay, early onset epilepsy (mean age of onset 24.6 ± 8.0 months) and varying severities of intellectual disability. Five patients with MRIs show hypoplasia or agenesis of the corpus callosum. Congenital cardiac abnormalities were also observed in four patients. Interestingly, congenital heart defects have been reported in patients with pathogenic variants in ATP6V1A and ATP6V1E1.³

ATP6V0C is an evolutionarily constrained gene as reflected by the high degree of amino acid homology between the human and yeast orthologues (72%), human and worm orthologues (63–67%), human and Drosophila orthologues (78%) and the low number of missense variants in gnomAD (n = 21 compared to an expected 108.5).³¹ Of the 18 unique patient missense variants, 16 are located in TM domains, with nine in TM4 that encompasses the p.E139 residue required for proton transport by the V-ATPase.¹ Consistent with evolutionary constraints on TM4, only one variant in gnomAD is located in this region of the protein (Fig. 2C).

Interestingly, four patients (Patients 4, 9, 15 and 19) were found to be mosaic for their identified ATP6V0C variant. On the basis of the available clinical information, the seizure phenotype of these patients may be less severe than for those with a germline variant (Table 1). Patient 4 (p.M53R) has seizure-like episodes that started at 14 months, but are not supported electrographically, Patient 15 (p.I132N) had seizure onset at 12 years of age and Patient 19 (p.G142D) has not reported any seizure or seizure-like episodes. In contrast to the other mosaic patients, Patient 9 (p.A95T) had seizure onset at 10 months. p.A95T also resulted in decreased V-ATPase activity in yeast (Fig. 5) and sterility in homozygous worms (Supplemental Table 4). The timing of the post-zygotic mutation event, and the affected tissues, can have a large influence on phenotypic presentation and severity in patients with somatic mosaicism, underlying differences in clinical presentation between patients with germline and mosaic variants.⁵⁷ Previous work has demonstrated the utility of identifying somatic mosaicism for clinical and genetic counselling outcomes in patients.⁵⁸

Nine copies of ATP6V0C and one copy of ATP6V0B assemble to form the intramembrane c-ring of the V-ATPase, which uses a rotary mechanism to translocate protons across the membrane (Fig. 1).²¹ Normally, frameshifting variants are predicted to cause NMD of the mutant mRNA, which would result in reduced protein levels. However, two of the four frameshift variants (c.340_355del16 and c.352_362delins) are located in last exon of ATP6V0C and are thus expected to escape NMD. The two frameshift variants in exon 2 may also escape NMD given the proximity of c.134_135delCT to the start codon and c.214delG being within 50 bp of the last exon-exon junction (Fig. 2A).⁵⁹ Additionally, modelling showed that outward-facing missense variants might act as a ‘stone in the gear’ between ATP6V0C and ATP6V0A inhibiting the rotatory mechanism, consistent with a dominant negative effect and the mechanism of action of V-ATPase inhibitors such as bafilomycin and archazolid (Fig. 4A and B).^45,60 Therefore, we speculate that missense variants and those predicted to escape NMD act via a dominant negative mechanism, while nonsense variants and microdeletions containing ATP6V0C act via haploinsufficiency as demonstrated by our Drosophila data (Fig. 3). However, additional studies, including the quantification of mRNA levels associated with each variant, are needed to delineate disease mechanisms more clearly.

Twelve disease-associated missense variants were examined in yeast. When the uptake of LysoSensor was measured, we saw that nine variants were associated with little to no fluorescence, indicating significant reduction or loss of V-ATPase activity. Higher levels of fluorescence were seen when the p.G63A, p.F137L and p.L150F patient variants were expressed (Fig. 5A and Supplementary Fig. 4). Growth curves generated by yeast expressing the 12 patient variants mirrored observations from the LysoSensor assay (Supplementary Fig. 5), with most variants resulting in little or no growth and intermediate levels of growth observed with the p.G63A, p.V74F, p.F137L and p.L150F variants (Fig. 5F). To further examine the effect of these intermediate variants on developmental and neurological function, we modelled p.G63A, p.F137L and p.L150F in worms. Expression of all three variants resulted in morphological delay as indicated by reduced body size and length at Day 1 of adulthood (Fig. 6A and B). Mutant worms also exhibited greater levels of paralysis and decreased lifespan when compared to N2 worms, and these phenotypes were exacerbated under osmotic stress (Fig. 6E–J). Mutants also exhibited increased activity and wave initiation rates, but speed was unaltered, suggestive of hyperactive and uncoordinated movement (Fig. 7A–C). The p.A95T variant that showed almost no V-ATPase function when tested in yeast (Fig. 5), resulted in homozygous sterility in worms, suggestive of a greater impact on V-ATPase function.

Additionally, we functionally examined, in yeast, three variants from gnomAD (p.R48W, p.G103S, p.M131I) that had high CADD scores and were predicted to be damaging by SIFT or Poly-Phen2, which contrasts the assumption that population variants would be benign (Supplementary Table 5). Expression of p.R48W and p.M131I resulted in growth that was more similar to the wild-type rescue compared to patient variants at 5 mM CaCl₂ and did not significantly differ from wild-type rescue as higher concentrations of CaCl₂ were tested (Fig. 5F and Supplementary Fig. 6). p.G103S, which had the highest CADD score of the gnomAD variants tested, resulted in significantly decreased growth compared to the wild-type rescue across all CaCl₂ concentrations tested (Fig. 5F and Supplementary Fig. 6). Nevertheless, p.G103S still yielded a significantly larger eAUC when compared to the best growing patient variant (p.F137L), suggesting the possibility of a level of decrease in V-ATPase activity that can be tolerated. However, we cannot exclude the possibility that mild clinical features could be associated with p.G103S or other rare gnomAD variants. Additional testing of population variants could establish the minimum level of V-ATPase activity required to maintain normal function. Such information would also aid in variant classification when novel patient variants are identified in the future.

The cellular mechanisms that explain the clinical features in patients with pathogenic variants in ATP6V0C are not yet known. However, given the importance of acidification driven by the V-ATPase in many cellular processes, it is likely that several pathways are impacted. First, the acidification of SVs within the central nervous system by the V-ATPase allows transporters, such as VGLUTs and VGAT, to load their respective cargo.^61,62 Our modelling data predict that outward-facing patient variants would interfere with the interaction between ATP6V0C and ATP6V0A (Fig. 4). Interestingly, patients with pathogenic ATP6V0A1 variants (encoding ATP6V0A) present with developmental delay, epilepsy, intellectual disability and cerebellar atrophy, similar to patients with pathogenic ATP6V0C variants.¹⁵ Recent work has shown that primary cortical neurons from Atp6v0a1^A512P/A512P mice, modelling a variant identified in a patient, have decreased SV neurotransmitter content and form fewer synapses.¹⁵ Given the functional interaction between ATP6V0C and ATP6V0A1, we speculate that synaptic defects probably contribute to disease pathology in patients with pathogenic ATP6V0C variants. These defects may also be independent of the ATPase activity of the V₁ domain as the V₀ domain is involved in neurotransmitter release independent of its acidification of SVs.^63–65

The V-ATPase also plays an important role within the trans-Golgi network (TGN) and inhibition of V-ATPase driven acidification can lead to improper trafficking and sorting of various membrane bound proteins, including neuropeptides and neuropeptide receptors.^66–68 Additionally, pathogenic variants in SLC9A7 in patients with developmental delay, intellectual disability and muscle weakness have been shown to cause alkalization of the TGN.⁶⁹ Therefore V-ATPase dysfunction may also lead to altered synaptic signalling via disruption of trafficking and sorting of receptors to the synapse. Consistent with this prediction of altered synaptic signalling, worms expressing the p.G63A, p.F137L and p.L150F variants displayed greater sensitivity to aldicarb and higher rates of paralysis compared to N2 worms (Fig. 7D).

Last, perturbations of lysosomal and autophagy pathways may also contribute to the epilepsy and neurodevelopmental phenotypes seen in these patients. The V-ATPase plays an essential role in acidifying endosomes, lysosomes and autophagosomes that ultimately creates the environment needed for proper trafficking and maturation of endocytic organelles and acid hydrolase function within the autophagy pathway.^2,70 Previous work by Nakamura et al.⁷¹ demonstrated that the V-ATPase is required for protein degradation from autophagic bodies in yeast vacuoles and Fassio et al.¹⁴ showed impairments in autophagic flux caused by pathogenic variants in ATP6V1A. In recent years, the contribution of impaired autophagy to neurodegenerative and neurodevelopmental disorders, including epilepsy, has risen in importance.^14,72,73 We hypothesize that the epilepsy and other neurodevelopmental phenotypes seen in patients with pathogenic V-ATPase variants may be due to a combinatorial effect of impaired synaptic signalling, trafficking and sorting of various membrane bound proteins, and defects along the endomembrane system including the lysosomal/autophagy degradation pathway. Further functional studies will be required to more thoroughly understand the mechanisms by which ATP6V0C variants lead to disease.

The prevalence of neurodevelopmental disorders, including epilepsy, resulting from variants in ATP6V0C is probably underestimated as this gene, to the best of our knowledge, is not currently included on commercially available epilepsy or intellectual disability gene panels. Of the 23 genes that encode for a subunit of the V-ATPase, 12, including ATP6V0C, are associated with disease. Ten additional members of the complex are expressed in the central nervous system but are currently not associated with disease (Supplemental Table 1).⁷⁴ Screening of these genes for potential pathogenic variants in patients with disorders such as epilepsy and intellectual disability should be undertaken. Additionally, longitudinal studies in patients with ATP6V0C mutations, and the identification of additional patients, will play an important role in resolving the full spectrum of co-morbidities associated with altered ATP6V0C and V-ATPase function.

In summary, we report 27 patients with heterozygous ATP6V0C variants who presented with developmental delay, early onset epilepsy and intellectual disability. In silico modelling suggests that most patient missense variants disrupt the interaction between the ATP6V0C and ATP6V0A subunits, and functional testing revealed that these variants decrease V-ATPase activity in yeast, and impair motor function, growth and lifespan in worms. Further work is needed to fully elucidate the mechanism(s) by which altered ATP6V0C function lead to the range of observed clinical phenotypes, and whether other V-ATPase subunits not currently known to cause disease harbour pathogenic variants in patients with neurodevelopmental disorders without a current genetic diagnosis.

Funding

This study was supported by a training grant appointment to K.A.M. (5T32GM008490), and an Emory University Research Council grant to A.E. J.B. is funded by a senior clinical investigator fellowship of the FWO-Flanders. This study was also supported in part by the Emory University Integrated Cellular Imaging Core (EICIC) and the Emory Integrated Genomics Core (EIGC) shared resources of Winship Cancer Institute of Emory University, National Cancer Institute (P30CA138292), the Canadian Rare Disease Models and Mechanisms Network (RDMM), Citizens United for Research in Epilepsy (339143), the Telethon Undiagnosed Diseases Program (TUDP; GSP15001), the Broad Institute of MIT and Harvard Center for Mendelian Genetics (Broad CMG) funded by the National Human Genome Research Institute, National Eye Institute and the National Heart, Lung and Blood Institute (UM1-HG008900 and R01-HG009141) and the National Institute for Neurological Disorders and Stroke (U01-NS077303-04S1 and R01-NS058721). The content is solely the responsibility of the authors and does not necessarily reflect the official views of the National Institutes of Health. This study was in part generated within the European Reference Network ITHACA, and uses data shared through RD-Connect, funded by the European Union (FP7/2007-2013, no. 35444).

Additionally, this research was made possible through access to the data and findings generated by the 100 000 Genomes Project. The 100 000 Genomes Project is managed by Genomics England Limited (a wholly owned company of the Department of Health and Social Care). The 100 000 Genomes Project is funded by the National Institute for Health Research and NHS England. The Wellcome Trust, Cancer Research UK and the Medical Research Council have also funded research infrastructure. The 100 000 Genomes Project uses data provided by patients and collected by the National Health Service as part of their care and support.

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.