Proteomic Investigation of Differential Interactomes of Glypican 1 and a Putative Disease-Modifying Variant of Ataxia

Abstract

In a currently 13-year-old girl of consanguineous Turkish parents, who developed unsteady gait and polyneuropathy at the ages of 3 and 6 years, respectively, we performed whole genome sequencing and identified a biallelic missense variant c.424C>T, p.R142W in glypican 1 (GPC1) as a putative disease-associated variant. Up to date, GPC1 has not been associated with a neuromuscular disorder, and we hypothesized that this variant, predicted as deleterious, may be causative for the disease. Using mass spectrometry-based proteomics, we investigated the interactome of GPC1 WT and the missense variant. We identified 198 proteins interacting with GPC1, of which 16 were altered for the missense variant. This included CANX as well as vacuolar ATPase (V-ATPase) and the mammalian target of rapamycin complex 1 (mTORC1) complex members, whose dysregulation could have a potential impact on disease severity in the patient. Importantly, these proteins are novel interaction partners of GPC1. At 10.5 years, the patient developed dilated cardiomyopathy and kyphoscoliosis, and Friedreich’s ataxia (FRDA) was suspected. Given the unusually severe phenotype in a patient with FRDA carrying only 104 biallelic GAA repeat expansions in FXN, we currently speculate that disturbed GPC1 function may have exacerbated the disease phenotype. LC–MS/MS data are accessible in the ProteomeXchange Consortium (PXD040023).

Introduction

Many rare diseases that manifest as Mendelian disorders are caused by pathogenic variants in a single gene. Although identification of the genetic cause of a given disease has improved significantly in the recent years, mainly due to better technologies, large amounts of data availability, and lower costs of short- and long-read next-generation sequencing (NGS) such as whole exome/genome sequencing (WES/WGS), it can still be challenging.¹ A key reason for this is that, despite a genetic diagnosis based on the detection of a specific pathogenic variant in a disease-associated gene, various factors such as disease onset, severity, and prognosis can vary widely due to environmental factors or genetic modifiers. Identifying such genetic modifiers can be a challenging task, mainly due to a limited amount of available whole genome data and the heterogeneity of the genomic backgrounds of individual patients.²

Using WGS in a patient with early-onset ataxia and polyneuropathy, we identified a highly deleterious homozygous missense variant in glypican 1 (GPC1) (MIM *600395), which we considered as a plausible disease candidate gene. To date, GPC1 has not been associated with a neuromuscular disorder or any other disorder. However, Gpc1 knockout in mice leads to aberrant brain and cerebellum morphology.³

GPC1 is a member of the glypican protein family, belonging to the large group of glycoproteins named heparan sulfate proteoglycans (HSPGs), characterized by their heparan sulfate (HS) modifications.⁴ The group of HSPGs consists of 17 family members, often classified based on their subcellular localization. These include secretory vesicles, the extracellular matrix (ECM), or the plasma membrane. Due to their negatively charged HS chains, membrane-bound HSPGs bind a vast variety of cytokines, signaling molecules, and receptors. Furthermore, they mediate several cellular processes such as morphogen gradients, ECM attachment, intracellular signaling, and cell migration.⁵

Glypicans belong to the group of membrane-bound HSPGs. In mammals, six glypican genes (GPC1 to GPC6) have been identified so far. They all share similar post-translational modifications (PTMs), including N-linked glycosylation, O-linked glycosylation (HS chains), and a glycosylphosphatidylinositol (GPI) anchor at their C-terminus, which is crucial for their attachment to the plasma membrane.⁵ Glypicans localize mainly to lipid rafts at the extracellular side of the plasma membrane and mediate developmental processes as well as stemness by regulating FGF, WNT, HH, and BMP signaling.^4,6−10

GPC1, the first member of the glypican family, is abundant in several tissues ranging from skin keratinocytes¹¹ to osteoblasts,¹² with high protein levels in the brain and cerebellum. It is a regulator of FGF2 signaling, stabilizing the complex of FGF2 and its receptor FGFR.¹³ A number of recent studies also emphasized the role of GPC1 in the nervous system, where it was detected in human neural stem cells (NSCs), and its abundance increased upon differentiation into cells of the neuronal lineage. In the same study, GPC1 has also been detected in cells of the astrocytic lineage but at lower levels.¹⁴ Furthermore, GPC1 mediates myelination of dorsal root ganglia (DRG) neurons in vitro in cocultures with Schwann cells, which is facilitated via interaction with collagens of the extracellular matrix.¹⁵ These findings underline potential roles of GPC1 for mature neurons.

Further studies in mice have shown that Gpc1 is coexpressed with the other glypicans (Gpc2–6) in the developing embryonic brain¹⁶ and has a role in the development of the central nervous system (CNS). Gpc1 knockout mice present a reduced number of neurons in the brain and show impaired cerebellar development. This is due to diminished FGF17 signaling, which is mediated by the loss of GPC1, resulting in premature differentiation of neurons and supporting the involvement of GPC1 in CNS development.³ A recent study has linked Dlp, a glypican ortholog in Drosophila melanogaster (D. melanogaster), to synaptic bouton growth at the neuromuscular junction (NMJ) and starvation-induced larval locomotion, providing further evidence for the role of glypicans in neuronal function.¹⁷ These studies in animal models provide a valuable source of a glypican’s role in neuronal development and synaptic processes and provide evidence for its involvement in NMJ function. However, currently, there is limited information on the role of GPC1 in the human nervous system.

To investigate the potential impact of the identified missense variant at the protein level, we probed the GPC1 interactome using coimmunoprecipitation (co-IP) and mass spectrometry (MS)-based proteomics. We used HEK293T cells transfected with wild-type (WT) versus mutant GPC1 protein expression constructs to identify candidate proteins with altered protein–protein interaction (PPI) levels. Significantly regulated proteins included members of the V-ATPase and mTORC1 complexes, which have strong potential to influence disease severity. Importantly, these proteins represent previously unknown interaction partners of GPC1.

Methods

Patient and Family Material, Informed Consent, and Ethical Approval

From the patient and her parents, blood samples and skin biopsies were obtained. From the latter, primary fibroblast cell lines were established according to standard protocols. Informed consent for sharing clinical data, genome data, the video of patient’s examination, and storing of the data in the EGA database were provided. The study was approved by the ethical committee of the University Hospital of Cologne under the number 13-022.

DNA Whole Genome Sequencing and Sanger Sequencing

WGS of the patients’ genomic DNA was performed as previously described.¹⁸ In brief, the DNA library for WGS was prepared using Illumina’s TruSeq PCR-Free sample preparation kit according to the manufacturer’s instructions. Using a Covaris E220 device, 1 g of genomic DNA was first fragmented to an average target size of 350–450 base pairs. Next, end repair, 3′ adenylation, and ligation of indexed sequencing adaptors were performed. To achieve the best cluster densities, insert sizes, and library diversities, the quality and concentration of each sequencing library were evaluated using a LabChip GX nucleic acid analyzer (PerkinElmer) followed by additional quality control using an Illumina MiSeq sequencer. Using the Illumina cBot, sequencing libraries were hybridized to the surface of HiSeqX flow cells (v2 or v2.5). Paired-end sequencing by synthesis was performed on Illumina HiSeqX sequencers using 2 × 150 cycles of incorporation and imaging. Real-time analysis included real-time conversion of image data into base calls. All steps in the sample preparation and sequencing workflow were monitored using an in-house laboratory information management system with barcode tracking of all samples and reagents. The GRCh37 reference sequence was used for read mapping. Raw sequences were aligned to the reference sequence using BWA version 0.7.10. Sequences in the BAM files were aligned to indels using GATKLite version 2.3.9 and a public set of known indels. PCR duplicates were marked using the Picard tool version 1.117. Variants with low map quality, low read depth (<6×), low Phred scaled SNP quality score (<20), and strong strand bias were filtered out. Variants were analyzed using the Clinical Sequence Miner application (deCODE Genetics). Low-quality variants, variants with call ratios of <20%, and variants with global allele frequencies of >0.01 were excluded from the analysis according to the Genome Aggregation Database (gnomAD). Variants were analyzed as present or homozygous in the affected individual. Repeat expansion analysis was performed using ExpansionHunter software. Variants were prioritized according to the in silico pathogenicity score using the Phred scaled score provided by Combined Annotation Dependent Depletion (CADD).

The following PCR primers were used for GPC1 variant verification by Sanger sequencing: GPC1-exon3-F: 5′-CTCAGAGCTGAGTCTCCCG-3′, GPC1-exon3-R: 5′-CCCAGGCAGTCCAGGTAGTC-3′.

Cell Culture

HEK293T and HeLa cells, as well as patient-derived primary fibroblasts, were maintained at 37 °C, 5% CO₂, in DMEM (Dulbecco’s modified Eagle medium) supplemented with 10% fetal bovine serum (Biochrom), 1% penicillin/streptomycin (Thermo Fisher), and amphotericin B (Promocell). Four different siRNAs targeting human GPC1 were purchased from Qiagen (GeneSolution siRNA, cat. no. 1027416) and used to transfect HEK293T cells in order to determine GPC1 antibody specificity (Figure S1). GPC1 expression plasmids were transfected using Lipofectamine 2000 (Thermo Fisher), and siRNA transfections were performed using Lipofectamine 3000 (Thermo Fisher) and DharmaFECT 1 (Horizon Discovery) following the manufacturers’ protocols.

Plasmids and Site-Directed Mutagenesis

Plasmids were generated by subcloning HA-tagged human GPC1 cDNA into the pcDNA3.1-CMV expression vector. The HA-tagged GPC1 cDNA construct was generated by gene synthesis (Genewiz). GPC1 c.424C>T was introduced using a QuikChange site-directed mutagenesis kit (Agilent) following the manufacturer’s protocol.

Coimmunoprecipitation

HEK293T cells were transfected with pcDNA3.1 plasmids containing N-terminally HA-tagged GPC1 WT or GPC1 c.424C>T. Cells were harvested 48 h after transfection by scraping in ice-cold lysis buffer (1% Triton, 150 mM NaCl, and 50 mM Tris–HCl, pH 8.0). The lysis buffer was supplemented with protease inhibitors (Roche) and phosphatase inhibitors (Thermo Fisher). Lysates were homogenized by passing through a 26 gauge needle and centrifuged at 4 °C and 10,000 rpm for 10 min. Lysates were then incubated with μMACS anti-HA MicroBeads (Miltenyi Biotec) followed by co-IP. For co-IPs, 2 mg of protein was combined with 50 μL of anti-HA MicroBeads in a total volume of 1 mL and incubated for 2 h at 4 °C on a rotator at 5 rpm. Magnetic separation was performed on μ-columns (Miltenyi Biotec) using the respective magnetic racks. After addition of samples, on column washes were performed four times using 400 μL of lysis buffer, and one final wash using 150 μL of 20 mM Tris–HCl (pH 7.5). After removal of the column from the magnet, proteins were eluted in 1% SDS, 5 mM DTT, and 20 mM Tris (pH = 8) for MS experiments or in Laemmli buffer (5% β-mercaptoethanol, 10% glycerol, 2% SDS, 60 mM Tris-Cl (pH 6.8), and 0.02% bromophenol blue) for Western blot analyses to a final volume of 80 μL. All co-IPs were performed from three biological replicates and utilized from different passage numbers of HEK293T cells and independent transfections.

MS Sample Preparation

Proteins were reduced with dithiothreitol (DTT) (5 mM final concentration) at 56 °C for 45 min and alkylated with acrylamide¹⁹ (20 mM final concentration) at room temperature (RT) for 30 min, and the reaction was quenched by further addition of 5 mM DTT and incubation for 10 min at RT. Subsequently, samples were processed using the single-pot, solid-phase-enhanced sample preparation (SP3) approach.²⁰ In brief, paramagnetic beads (Sera-Mag magnetic carboxylate-modified particles, nos. 44152105050250 and 24152105050250, Cytiva) were prepared at a 1:1 (w/w) ratio and added to the samples, and the acetonitrile (ACN) content of the solution was adjusted to 75% (final bead concentration: 1.5 μg/μL). Samples were incubated at 1000 rpm in a thermomixer for 20 min at RT, tubes were placed in a magnet, the supernatant was removed, and particles were washed twice with 100 μL of 80% ethanol and once with 100 μL of ACN. Subsequently, beads were resuspended in 100 mM NH₄HCO₃, and proteins were digested with trypsin (Promega) at a 1:50 (w/w) protease-to-protein ratio at 37 °C overnight. The next day, peptides were recovered by magnetic immobilization of the beads and transfer of the peptide-containing supernatant to a fresh tube. Samples were acidified with 0.1% acetic acid (final concentration), and peptides were desalted using C₁₈ StageTips²¹ packed with 3 discs of C₁₈ material (Empore 2215, 3 M). Eluate fractions were dried in a vacuum centrifuge and stored at −20 °C until MS analysis.

LC–MS/MS Data Acquisition

Dried peptides were resuspended in 5% ACN and 5% formic acid (FA), and 30% of the sample was combined with 50 fmol of iRT-peptides.²² Samples were analyzed by LC–MS/MS using an ultrahigh-performance nanoliquid chromatography system (Dionex Ultimate 3000 nano-UHPLC) coupled to an Orbitrap Fusion Lumos mass spectrometer (both Thermo Fisher). Peptides were loaded directly to a 40 cm in-house manufactured analytical column. Columns were produced from spray tips generated from 360 μm outer diameter, 100 μm inner diameter fused silica capillaries with a P-2000 laser puller (Sutter Instrument) followed by their packing with 3 μm ReprosilPur AQ C₁₈ particles (Dr. Maisch). After sample loading for 25 min at 600 nL/min with 99% solution A (0.1 %FA) and 1% solution B (90% ACN, 0.1% FA), peptides were eluted using a 60 min linear gradient from 1 to 35% solution B at a flow rate of 300 nL/min. Data were acquired in the data independent acquisition (DIA) mode with peptide precursor isolation windows of m/z 35.4 (0.5 m/z overlap), covering a range of m/z 368–1182.5 at a cycle time of 3 s. Precursor ions were fragmented by higher collisional dissociation (HCD) with a normalized collision energy of 27%, and MS2 spectra were analyzed in an Orbitrap mass analyzer at a resolution of 30,000, an AGC target setting of 1 × 10⁶, and a maximum ion injection time of 60 ms. MS1 scans were acquired at a range of m/z 350–1200 and a resolution of 120,000 in an Orbitrap mass analyzer with an AGC target setting of to 5 × 10⁵ and a maximum injection time of 20 ms.

Data Analysis

DIA raw data were analyzed with Spectronaut 14.7.201007 (Biognosys) using the Pulsar search engine following the direct DIA approach. For spectral library generation, DIA data were matched against UniProt Homo sapiens (74,416 entries, release date 06-2019) in combination with the MaxQuant contaminant database (246 entries, release 04-2019) and sequences of the 11 iRT peptides.²² Enzyme specificity was set to trypsin/P with up to two missed cleavage sites, propionamide (Cys) was selected as fixed modification, and acetylation (protein N-term) and oxidation (Met) were defined as variable modifications. For data extraction and scoring at the MS1 and MS2 level, ideal mass tolerances were set by Spectronaut (dynamic mode), and the most intense 3 to 6 fragment ions (b and y) in the range from m/z 300 to 1800 were included. For peptide identification, a false discovery rate (FDR) threshold of 1% was applied on the peptide spectral match (PSM), peptide, and protein level. Normalized retention times were generated based on iRT peptide values (autoassign iRT source). DIA data were matched to the spectral library using dynamic assignment of mass tolerances and selection of the traces with the highest intensity at both the precursor and fragment ion level. For peptide and protein identification, the following cutoffs were applied: precursor PEP value: 1; precursor Q value: 0.01; protein Q value (experiment): 0.01; protein Q value (run): 1. Peptide quantification was based on the mean area of the three most intense fragment ions with enabled interference correction. Label-free protein quantification was performed from up to 3 peptides per protein, utilizing their mean quantity with the Quant 2.0 protein quantification logic. Data were exported from Spectronaut to *.txt files, and further data processing was performed with Perseus version 1.6.14.0.²³ Volcano plots were generated in R (version 4.2.2) with the EnhancedVolcano package (version 1.16.0). Gene ontology (GO) analysis was performed via PANTHER 17.0 (http://www.pantherdb.org/), the complete unprocessed list of identified proteins was uploaded as a background (Table S4), and the GPC1 WT interactome (Table S5) was uploaded for the GO enrichment analysis against the background.

Western Blotting and Antibodies

Western blotting was performed following standard protocols with 0.2 μm nitrocellulose membranes. For both blocking of membranes and antibody incubation, 5% skim milk and 1% BSA in TBS-Tween (0.5%) were used. The following antibodies were used: α-GPC1 (1:2000, Sigma: SAB2700282), α-MOGS (1:1000, Proteintech: 17859-1-AP), α-GNA13 (1:1000, Proteintech: 67188-1-Ig), α-CANX (1:1000, Proteintech: 10427-2-AP), LAMTOR2 (1:1000, Cell Signaling: 8145), α-HSP90 (1:1000, Cell Signaling: C45G5), ATP6V1F (1:1000, Proteintech: 17725-1-AP), α-EGFR (1:1000, Santa Cruz: sc-03), mouse α-β-actin-HRP (1:5000, Proteintech: HRP6008), α-rabbit-HRP (1:2500, Cell Signaling: 7074P2), and α-mouse-HRP (1:2500, Dianova: 115-035-146).

Immunocytochemistry and Antibodies

Twenty thousand HEK293T cells were seeded on coverslips in a well of a 12 well-plate a day before transfection. They were transfected with HA-tagged GPC1 expressing plasmids, and as a control, an empty HA plasmid was used. Forty-eight hours after transfection, cells were fixed in 4% PFA and permeabilized using 0.2% Triton X-100 in PBS for 10 min and blocked in blocking buffer for 1 h (5% BSA and 0.2% Tween-20 in PBS). Samples were incubated in primary antibodies overnight in blocking buffer. The next day, coverslips were washed with PBS and incubated with fluorophore-coupled secondary antibodies for 1 h in blocking buffer prior to mounting using ProLong Gold (Thermo Fisher: P36935). Microscopy analyses were performed using a Zeiss AxioImager with Apotome.2 attachment. Images were acquired at 63× magnification with oil immersion. The following antibodies were used for immunocytochemistry analyses: α-GM130 (1:200, Abcam: ab52649), α-CANX (1:200, Proteintech: 10427-2-AP), and α-GPC1 (1:2000, Sigma: SAB2700282).

Subcellular Fractionation

HEK293T cells were harvested 48 h after transfection in ice-cold PBS using a cell scraper, transferred to microtubes, and centrifuged at 500 g, and the supernatant was discarded to obtain a cell pellet. Cell pellets were immediately processed for subcellular fractionation using a subcellular protein fractionation kit for cultured cells (Thermo Fisher, catalog no. 78840) following the manufacturer’s protocol.

Results

GPC1 R142W as a Putative Disease-Modifying Mutation in a Friedreich’s Ataxia Patient

A female patient, who is currently 13 years old, was extensively clinically investigated at the age of 8 years. She presented with motor developmental delay and frequent falling (Video S1). Prenatal, natal, and postnatal histories were uneventful. Early developmental milestones were normal. Parents were first cousins, and the family history was otherwise negative (Figure 1a). The past medical history revealed that she first presented at the age of 3 years with frequent falls, unsteady gait, and difficulties to get up from the floor. The course was progressive in nature, she began to have hand tremors, and an electromyography study at the age of 6 years was compatible with a sensory neuropathy. A targeted gene panel for neuromuscular disorders¹⁸ was negative. She was initially diagnosed with chronic inflammatory demyelinating neuropathy (CIDP) and received immunotherapy. By the age of 9 years, she was only able to sit with support, developed thoracal scoliosis, and lost independent ambulation. Cranial magnetic resonance imaging (MRI) was normal.

Figure 1.

GPC1 R142W as a putative disease-modifying mutation in Friedreich’s ataxia. (a) Pedigree represents the affected index patient and her affected brother (II.1 and II.2) as well as the unaffected parents (I.1 and I.2). FRDA-affected individuals are indicated in black. GPC1 c.424C>T genotypes are indicated under each confirmed individual. (b) Confirmation of the GPC1 c.424C>T variant by Sanger sequencing, detected as homozygous for the index patient and as heterozygous for the parents. (c) Schematic representation of GPC1 and its post-translational modifications. The GPI anchor is required for membrane attachment, and three C-terminal O-linked glycosylation sites are represented in red and two N-linked glycosylation sites in gray. Created with BioRender.com. (d) Conservation of arginine at position 142 within various species shown by amino acid alignment. (e,f) Western blot analysis of GPC1 levels in patient-derived fibroblasts. A significant increase in intensity for the band above 250 kDa was observed for the patient compared to the parents. Student’s two sample t-test; ****p < 0.0001, **p < 0.01, ns.: not significant (≥0.05).

For further genetic diagnosis, we performed a hypothesis-free short-read WGS, when the child was 8 years old. Due to the 145 Mb length of the runs of homozygosity (ROH) confirming parental consanguinity, we focused on the homozygous variants located in ROH regions larger than 0.5 Mb. Sparing the variants that are nonsynonymous with an allele frequency of <1% and manually excluding technical artifacts resulted in a final candidate list of nine variants (Table S1). We further excluded variants with a low in silico score, nonrelevant cellular function, and/or tissue expression outside the central and peripheral nervous system. Finally, we identified a homozygous variant in GPC1 (ENST00000264039: c.424C>T, p.R142W), located on a 1 Mb ROH on chromosome 2. The heterozygote carrier status of the parents was later confirmed by Sanger sequencing (Figure 1b). According to the gnomAD database, the variant was previously observed in six alleles as heterozygous and never as homozygous (allele frequency: 2 × 10^–5). Furthermore, the variant was identified in one somatic tumor sample (gastric cancer) among 585 tested donors (https://dcc.icgc.org/mutations/MU128862589). The variant is predicted as deleterious in silico (CADD score 26.3). Based on the involvement of GPC1 in the brain and particularly in cerebellar and NMJ development as shown in mice and D. melanogasterGpc1 knockout models,^3,8 we assumed that GPC1 might be the putative causative gene for this early neuropathy.

At the age of 10.5 years, the child developed a dilated cardiomyopathy and was referred to the pediatric neurology outpatient clinic. On neurological examination, the girl was wheelchair-bound, had dysarthria, dysmetria, and intentional tremor. She could only take a few steps with a steppage gait. She had marked thoracal scoliosis, kyphotic posture, and pes cavus deformity. Deep tendon reflexes were absent, and the Babinski sign was positive bilaterally. Furthermore, worsening of truncal ataxia upon eye closure with decreased vibration sense in the lower extremities was observed. Based on these additional symptoms, a suspected clinical diagnosis of Friedreich’s ataxia (FRDA, MIM *229300) was made and subsequently confirmed by genetic testing of the FXN gene (MIM *606829). Within a 12-month follow-up on idebenone treatment, the patient developed progressive left ventricular dysfunction and heart failure, with a decrease in the ejection fraction from 38 to 10–15%. For the past two months, her fasting glucose level was detected to be >300 mg/dL (N: 70–100 mg/dL), and she developed diabetes mellitus.

Based on the short-read WGS, the GAA repeat expansion in the FXN gene was missed. However, using recently developed algorithms to identify the repeat expansion reliably (ExpansionHunter 5.0), we could observe a homozygous 104 GAA trinucleotide repeat expansion from the short-read WGS data. The patient’s currently 5-year-old brother was also found to have a GAA repeat expansion in the FXN gene at a presymptomatic stage at the age of 3 years and has a mild unsteady gait without cardiac involvement.

Consequently, we revised our initial hypothesis and considered the GPC1 variant as a putative modifier that might further have influenced the FRDA disease severity. To study the potential impact of the GPC1 mutation on the protein’s interactome, we conducted co-IPs in combination with proteomic analysis and further functional interpretation of the results. To our knowledge, the patient’s mutation, GPC1 R142W, does not overlap with any of the known PTM sites (Figure 1c) of the protein or with any known function.

First, we investigated GPC1 protein abundance in primary patient-derived skin fibroblasts, using Western blot analysis. As controls, skin biopsy-derived fibroblasts from the parents were used. Overall, we did not observe any visible differences in the growth rates of the obtained primary fibroblast lines. The Western blot analysis revealed an increased band intensity for GPC1 above 250 kDa in the patient (Figure 1e,f). High-molecular-weight bands of GPC1 were reported to be due to differential O-glycosylations (HS chains), which can vary greatly in length.^24,25 The length and modification status (e.g., sulfation) of HS regulate its binding capacity to other molecules, such as growth factors, cytokines, and receptors.⁵ Therefore, an increased band intensity could reflect altered interaction dynamics of GPC1 R142W with other proteins. In order to investigate the GPC1 interactome and potentially altered PPIs due to the GPC1 R142W in further detail, we performed coimmunoprecipitations (co-IPs) in combination with MS-based proteomic analyses.

Investigation of the GPC1 Interactome by Mass Spectrometry-Based Proteomics

For GPC1 interactome analyses by MS, we aimed for a clean and reproducible system without the genetic heterogeneity typically presented by patients’ fibroblasts. Moreover, we wanted to exclude any impact from the mildly expanded FXN gene. For this purpose, we transiently expressed single HA-tagged GPC1 WT and R142W in HEK293T cells and performed anti-HA co-IPs from these cells (Figure 2a). Initially, we generated an N-terminally tagged GPC1 expression construct by insertion of an HA tag between the BM40 signal peptide, which is required for plasma membrane targeting, and the GPC1 core protein (Figure S2a). The originally predicted GPC1 N-terminal signal peptide was replaced by a BM40 signal, which was previously shown to retain cleavage functionality in a similar architecture.²⁶ This minimized the possibility of dysfunctional cleavage and thus erroneous subcellular localization of our GPC1 constructs. We investigated the expression of the HA-tagged GPC1 constructs in HEK293T cells by Western blotting, confirming that they result in banding patterns similar to endogenous GPC1, reaching to a molecular weight of >250 kDa (Figure S2b). In these analyses, we did not observe any differences in the expression levels of BM40-HA-GPC1 WT and BM40-HA-GPC1 R142W, which we will hereinafter refer to as GPC1 and GPC1 R142W, respectively.

Figure 2.

Investigation of the GPC1 WT and R142W interactome by MS-based proteomics. (a) Workflow for sample preparation, co-IP, and MS analysis. Created with BioRender.com. (b) Western blot analysis of the HA-GPC1 co-IP eluate and whole cell lysate input fractions. (c) Total numbers of protein IDs detected by LC–MS/MS in individual replicates.

Next, we performed co-IPs from whole cell lysates with anti-HA paramagnetic beads and confirmed successful GPC1 enrichment by Western blotting (Figure 2b). To investigate the interactomes of GPC1 and GPC1 R142W, co-IPs were performed in three independent biological replicates and the eluate fractions analyzed by DIA-LC–MS/MS. Furthermore, we included cells transfected with GFP as controls. This resulted in three datasets: GFP-transfected cells, allowing definition of background proteins that nonspecifically bind to the beads, GPC1 WT to identify the GPC1 interactome, and GPC1 R142W to analyze the impact of the patient’s mutation on GPC1 PPIs. A similar number of proteins were identified for both GPC1 IPs, giving a total of 2229 protein groups, of which 2206 were identified for GPC1 and 2224 for GPC1 R142W. We observed high reproducibility between individual replicates in terms of the number of proteins identified (Figure 2c) as well as their intensities (Figure S3a). Furthermore, the data obtained from GPC1 IPs showed a distinct separation from GFP-transfected control samples and high reproducibility between biological replicates, as determined by unsupervised hierarchical clustering (Figure S3a), principal component analysis (PCA, Figure S3b), and Pearson correlation analyses (Figure S4). Furthermore, by using principal component 4, a clear separation of GPC1 and GPC1 R142W was possible, indicating differences in their interactome (Figure S3b).

Proteomic Analysis of the GPC1 Interactome

First, we focused on the GPC1 interactome in our dataset and compared the signal intensities between GFP (mock transfected control group) and GPC1 analyses. After removal of protein groups with <3 valid values in any of the groups, we calculated the fold-change of the protein levels and p-values for GPC1 compared to GFP. We identified 198 significantly enriched proteins (≥2-fold upregulated, p < 0.05) compared to GFP, which present high confidence interaction partners of GPC1 (Figure 3a). These include proteins previously shown to interact with GPC1, such as FGF2,^13,27 confirming successful co-IP of GPC1 and its interaction partners.

Figure 3.

Proteomic analysis of the GPC1 interactome. (a) Volcano plot representing fold-change and p-values of the label-free quantification for GPC1 and GFP-transfected HEK293T cells. GPC1 interacting proteins passing the significance threshold (log₂(fold-change) ≥ 1, −log₁₀(p-value) ≥ 1.3) are highlighted. (b) Gene ontology (GO) enrichment analyses of the GPC1 interactome. (c–f) Subcellular distribution analysis of the GPC1 interactome. (c) ER-Golgi-, (d) PM and signaling-, (e) endosome-, and (f) lysosome-associated proteins are highlighted in the respective subplots. PM: plasma membrane.

Subsequently, we investigated the properties of the GPC1 interactome by a keyword search of the gene ontology (GO) enrichment results. GPC1 is a membrane-bound protein that is known to be associated with lipid rafts and interacts with other membrane (−associated) proteins.^7,28,29 In line with this localization, 180 of the 198 interacting proteins are associated with membranes, including 105 proteins that localize to the plasma membrane. We further observed an enrichment of ER- (62 proteins) and Golgi- (43 proteins) associated proteins, consistent with their role in GPC1 biosynthesis, post-translational modification, and vesicular transport. Interestingly, we furthermore found a strong enrichment of the endosomal (31 proteins) and lysosomal (36 proteins) compartments, which are related to GPC1 endocytosis and degradation. Next, we performed a GO analysis of the 198 proteins identified as the GPC1 interactome. Enrichment analysis of GO terms related to the “biological process” revealed significant enrichment of endolysosomal organization and signaling pathways, mTORC1 signaling, and ER stress and Golgi organization. Consistent with these findings, the V-ATPase complex, ER-Golgi transport vesicles, endolysosomal vesicles, and the membranes of these compartments were highly enriched for the GO term “cellular component” (Figure 3b). Here, we mainly focus on the GPC1 interactome at these subcellular locations, as they resemble the sites of GPC1 synthesis, post-translational modification, sorting, and recycling/degradation.

In the group of proteins associated with ER and Golgi, we have observed two main categories. The first group was involved in anterograde and retrograde vesicular trafficking between the ER, the Golgi, and the plasma membrane (Figure 3c). These include p24 protein family members (TMED1, TMED2, TMED9, and TMED10), which have been shown to be involved in COPI- and COPII-mediated vesicular trafficking between ER and Golgi and have been shown to play a role in recognizing GPI-anchored cargoes in yeast, mouse, and human cell lines.³⁰ Furthermore, we detected RAB1B, which is involved in Golgi and ER vesicular trafficking,^31,32 and BCAP31 that functions as an ER chaperone with additional roles in the anterograde transport of membrane proteins.^33,34

The second group of candidates are related to protein glycosylation. Most of these proteins have been shown to play a role in HS-glycosaminoglycan (GAG) biosynthesis (EXTL2, B3GALT6, and UXS1)³⁵⁻³⁹ and sulfation (HS2ST1 and SLC35B2).^40,41 Other roles include GPI biosynthesis in the ER (PIGG and PIGK) and quality control of glycoproteins, as well as two-step hydrolysis of N-linked glucose to asparagine (GANAB, an α subunit of glucosidase II).^42,43 Because these proteins are directly involved in PTMs on GPC1, they can be considered as putative modifiers.

GPC1 is a membrane HSPG and in line with our expectations, a number of candidates associated with the plasma membrane and signaling cascades were identified (Figure 3d). These candidates included some membrane HSPGs such as GPC3, GPC4, GPC6, and SDC2. Glypicans and syndecans can cooperate at the plasma membrane and have the ability to regulate and fine-tune pathways including but not limited to WNT, FGF and, HH signaling.^4,17,44,45 Having glypicans highly associated with WNT signaling, we identified CSNK1G1, CSNK1G2, and CSNK1G3 (Casein Kinase I subunits) as well as PLPP3, which are known regulators of this signaling pathway.⁴⁶⁻⁴⁸ Additionally, GPCRs (GPRC5C and GPR180), G protein regulators (RGS17 and RGS19), and several G proteins (GNA13, GNAI1, GNAI3, GNAQ, GNAS, GNB2, GNG5, and GNG12) (Table S5) were enriched in our dataset, further supporting GPC1 involvement in WNT signaling. With respect to FGF signaling, we found FGF2 and FRS2, both of which are FGFR ligands, as potential GPC1 interacting proteins. Furthermore, several MAPK signaling pathway members and/or regulators were enriched, such as RRAS2, KIDINS2020, LAMTOR1, LAMTOR2, LAMTOR3, and FYN, that may directly or indirectly interact/crosstalk with the above-mentioned signaling cascades.

Glypicans are mainly located at the plasma membrane, and GPC1 has been shown to be recycled via the endolysosomal pathway. After GPC1 internalization by caveolin-mediated endocytosis, it accumulates in RAB7-positive late endosomes, where its HS chains are modified and cleaved to generate anhydromannose (HS-anMan), which can interact with other molecules.^49,50 Consistent with the recycling pathway of GPC1, RAB35 and RAB10 were enriched in our dataset as recycling endosome markers.⁵¹⁻⁵³ In addition, we detected FLOT1 and FLOT2, which are located at the plasma membrane, caveolar membranes, endosomes, and lysosomes⁵⁴ and are ascribed major roles in vesicular trafficking.⁵⁵ Interestingly, we also identified components of the ESCRT-I (VPS28) and ESCRT-III (CHMP1A, CHMP2A, CHMP2B, CHMP4B, and CHMP5) complexes. The ESCRT complex is known to mediate the formation of multivesicular bodies (MVB), which are endosomal acidic vesicles⁵⁶ that eventually fuse with lysosomes to target ubiquitinated proteins for degradation (Figure 3e).^57,58

Regarding lysosomes, we detected several members of the vacuolar ATPase (V-ATPase) (Figure 3f), a MDa complex consisting of a membrane-embedded proton pumping V0 domain and a cytosolic ATP hydrolyzing V1 domain.⁵⁴ For both domains, we identified several subunits (V1 domain: ATPV1A, ATPVB2, ATPVC1, ATPVD, ATPVE1, ATPVF, ATPVG1, and ATPVH; V0 domain: AT6V0A1, AT6V0A2, AT6V0A3, AT6V0C, and AT6V0D1) as well as the accessory subunits ATP6AP1 and ATP6AP2, implying efficient enrichment of the intact complex. V-ATPases, whose subunits are highly enriched in our dataset, facilitate acidification/pH regulation of various intracellular compartments^59,60 and localize, e.g., to the plasma membrane, to the Golgi apparatus, and to endosomes and lysosomes. V-ATPases are functional when the V0 and V1 domains are assembled and inactive when they are reversibly dissociated,⁵⁹ indicating an interaction of the active complex with GPC1. With all members of the Ragulator complex (LAMTOR1, LAMTOR2, LAMTOR3, LAMTOR4, and LAMTOR5) as well as the associated RAG GTPases (RRAGA, RRAGB, RRAGC, and RRAGD), we identified another fully assembled lysosomal complex in our dataset. Ragulator is a key factor for mTORC1 activation at the lysosomal membrane, which also requires binding of the RAG GTPases as well as the V-ATPase.

To confirm the results of our GO analyses, we further investigated the subcellular localization of HA-tagged GPC1 and GPC1 R142W in HEK293T cells. Simultaneous immunostaining of the HA tag with the ER marker CANX (Figure 4a) and the Golgi marker GM130 (Figure 4b) revealed similar levels of ER and Golgi localization for GPC1 and GPC1 R142W. Furthermore, we performed subcellular fractionation experiments in combination with Western blot analysis, which revealed a significant enrichment of both GPC1 constructs in the membrane-enriched fraction (Figure 4c), further confirming our results by MS analyses. These results also demonstrate that the transiently expressed GPC1 constructs are quantitatively GPI-modified, resulting in correct membrane localization of all proteins.

Figure 4.

Subcellular localization of HA-tagged GPC1 WT and GPC1 R142W in transfected HEK293T cells. (a) Investigation of ER localization by coimmunostaining with CANX. (b) Investigation of Golgi localization by coimmunostaining with GM130. (c) Western blot analysis of subcellular cytosolic and membrane fractions. HA-tagged GPC1 WT and GPC1 R142W were exclusively detected in the membrane fraction. Phal.: phalloidin, No trans.: no transfection.

Given the significant enrichment of components of the endolysosomal pathway in our dataset, we also investigated our results for possible contamination of intact intracellular vesicles, which could originate from insufficient cell lysis. Analysis of our complete dataset (>2000 proteins) (Tables S2–S4) for the presence of core endosomal and lysosomal markers did not result in identification of large numbers of known lysosome-/endosome-related proteins such as LAMP1/2, LC3, ATG5/12, RAB4, and EEA1, further supporting the direct interaction of GPC1 with these detected complexes.

Comparison of the GPC1 vs GPC1 R142W Interactomes

To investigate the alterations in PPIs caused by the identified GPC1 mutation, we compared the interactomes of GPC1 and GPC1 R142W. We considered all proteins that were enriched by ≥1.5-fold in any of the datasets compared to GFP, resulting in 263 proteins (Figure S5 and Table S7). Subsequently, we directly compared the signal intensities of these proteins for GPC1 and GPC1 R142W and defined proteins with p < 0.05 and fold-change ≥ 1.5 as interacting differentially. In total, we identified 16 candidate proteins, of which four and twelve presented with decreased or increased interaction levels, respectively, for GPC1 R142W compared to GPC1 (Figure 5a,b).

Figure 5.

Comparison of the GPC1 and GPC1 R142W interactomes. (a) Volcano plot representing fold-change and p-values of the label-free quantification for GPC1- and GPC1 R142W-transfected HEK293T cells. Differentially interacting proteins passing the significance threshold (log₂(fold-change) ≥ 0.57 and ≤ 0.57, log₁₀(p-value) ≥ 1.3) are highlighted. (b) List of candidate proteins showing significant differences between both constructs. (c) Confirmation of selected MS candidates in co-IPs by Western blotting and (d) intensities based on densitometric quantifications of Western blot bands, normalized to the HA signal of the respective sample. ***p < 0.001, ns.: not significant (≥0.05).

Interestingly, five members of the interaction group showing increased values were members of the cytoplasmic V1 domain of the V-ATPase complex (ATP6V1A, ATP6VB2, ATP6VE1, ATP6VF, and ATP6VG1), which can reversibly dissociate from the membrane-embedded V0 domain, for regulation of complex activity. Another protein with increased interaction was LAMTOR2, a member of the Ragulator complex that is known to interact with the V-ATPase complex at the lysosomal membrane and involved in regulation of mTORC1 signaling.^61,62 In addition, we identified two Golgi-associated proteins with increased interaction levels: GORASP2, a Golgi-stacking protein, and HS2ST1, an HS 2-O sulfotransferase. Interestingly, HS2ST1 is responsible for the 2-O sulfation of HS chains, which can influence the ligand affinity and specificity of HSPGs.⁶³ Therefore, differential interaction of GPC1 with HS2ST1 could alter the 2-O sulfation status of GPC1 R142W, eventually resulting in an altered ability to mediate downstream signaling pathways. Finally, we found the ER-resident chaperone CANX as a candidate for different levels of interaction (Figure 5a–d). CANX is known to mainly bind misfolded N-linked glycosylated proteins,⁶⁴ and enhanced interaction with GPC1 R142W could be due to high levels of protein misfolding as a result of the missense mutation, ultimately influencing the functionality and localization of the protein. Combined with the observed colocalization of CANX with both GPC1 and GPC1 R142W (Figure 4a), these results suggest that the observed quantitative differences of GPC1 with ER proteins are indeed due to direct interaction between the individual proteins and not due to secondary effects caused by the differential subcellular distribution of the protein.

In total, four proteins with reduced levels of interaction were identified (Figure 5b). The ER-localized MOGS (glucosidase I) is involved in the N-glycosylation of proteins. Reduced GPC1-MOGS interaction could result in the altered GPC1 glycosylation state as it is known to bear two N-linked glycosylation sites. We further identified lower levels of GNA13, a G protein α subunit that is a downstream member of GPCR signaling. Decreased GPC1-GNA13 interaction could influence GPCR activity and cellular homeostasis. Finally, the E3 ubiquitin ligase TRIP12 and the nucleolar protein NOP9 showed slightly reduced levels of interaction.

We also examined six GPC1-interacting proteins that showed altered levels of interaction between GPC1 and GPC1 R142W by co-IP and Western blot analyses. These included GNA13, the ER proteins MOGS and CANX, and the lysosomal proteins ATP6V1F, ATP6V1B2, and LAMTOR2 (Figure 5c,d). While we could confirm the interaction of GPC1 and GPC1 R142W with all six proteins, we could observe significant differential regulation only for two of them (CANX and GNA13), and MOGS presented with a tendency to reduced interaction levels for all replicates. Intriguingly, we were able to show the inverse regulation of two ER proteins MOGS and CANX involved in protein N-glycosylation and binding to misfolded N-glycosylated proteins, respectively. This could represent a functional link between these proteins and GPC1, e.g., through reduced MOGS-dependent N-glycosylation resulting in more pronounced binding of CANX.

Discussion

In this study, we identified the biallelic GPC1 mutation R142W by WGS in a patient with an early-onset ataxia and neuropathy at 3 years of age, who developed dilated cardiomyopathy and kyphoscoliosis at 10.5 years of age and was diagnosed with Friedreich’s ataxia (FRDA). FRDA patients with full-penetrance disease-causing expanded alleles (66 to 1300 GAA repeats) have a mean age of onset at 10 to 15 years, although symptoms can rarely begin earlier.⁶⁵ In the context of the patient’s early onset of symptoms (3 years of age) and the absence of a known gene for a neuromuscular disorder in the WGS data, we initially considered the GPC1 as a potential novel disease-causing mutation, which we later reclassified as a potential genetic disease modifier upon identification of a pathogenic GAA repeat expansion in the FXN gene. Based on the rather modest increase of GAA repeats (n = 104), the severe phenotype is rather unexpected. It is well-known that FRDA disease severity correlates with an increased length of GAA repeats, and every 100 GAA repeats on the smaller repeat allele is associated with a 2.3-year (SE 0.2) earlier onset.⁶⁶

To investigate potential effects of the GPC1 variant at the molecular level, we investigated the GPC1 and GPC1 R142W interactomes by MS-based proteomics. This enabled the characterization of GPC1 PPIs and the identification of differentially interacting proteins, which could represent disease modifier signaling pathways that act either dependent or independent of FXN.

For the GPC1 interactome, we were able to identify proteins originating from all subcellular localizations where GPC1 has been shown to localize and validated six of them by independent co-IPs and Western blot analysis. This not only confirms the correct localization/distribution of our transiently expressed HA-tagged construct but also the successful co-IP and label-free quantification by MS compared to the GFP-transfected control cells. For example, we identified several proteins known to be involved in ER-Golgi vesicular transport and glycosylation at the plasma membrane, which also affect signal transduction, or those associated with the endolysosomal pathway. Since GPC1 is a membrane HSPG with multiple glycosylated residues, PPIs at the ER and Golgi may be involved in GPC1 biosynthesis, post-translational modifications such as glycosylation, and transport to the plasma membrane. Glypicans in the latter subcellular compartment have been shown to fine-tune multiple signaling pathways and play a role in, for example, nervous system development and myelination.¹⁵ Furthermore, alterations in these proteins have been identified as biomarkers in pancreatic, prostate, and colorectal cancer.⁶⁷ The PPIs identified in this study could shed light on how GPC1 (and glypicans in general) is involved in such diseases or contribute to a better understanding of disease pathologies and developmental processes with GPC1 involvement.

With respect to the recycling and degradation of GPC1, we mainly detected proteins located in two major lysosomal complexes, the V-ATPase complex and the Ragulator complex, for which in our co-IP experiments, we found the majority of subunits. It has been shown that the V-ATPase interacts with Ragulator and via this also with the RAG GTPases on the lysosomal membrane, which is crucial for the activation of mTORC1. Furthermore, V-ATPase activation by the assembly of its V1 and V0 domains is triggered by starvation-induced mTORC1 inhibition, leading to acidification of the lysosomal lumen to activate lysosomal enzymes and hence catabolic processes.⁶⁰ Whether the co-IP of GPC1 is due to a direct interaction with the V-ATPase and Ragulator and if such an interaction would be able to directly influence lysosomal nutrient sensing or if this observation is due to the presence of GPC1 in late endosomes/lysosomes as part of its recycling/degradation pathway warrants further investigation. Membrane HSPGs have been shown to be associated with the internalization of receptors and ligands, ultimately affecting downstream signaling pathway activity.⁵ Therefore, the direct or indirect interaction of GPC1 with V-ATPase and Ragulator could be a way to regulate mTORC1 signaling analogous to the processes observed at the plasma membrane. Consistent with these lysosomal protein PPIs, it is well-known that glypicans located at the plasma membrane can mediate sonic hedgehog (SHH) internalization and target it for degradation to regulate HH signaling.⁶⁸ Taken together, our data are indicative of distinct PPIs of GPC1 in the lysosomal compartment, rather than coincidental localization due to its degradation.

In this regard, several studies in the recent years demonstrated that lysosomes play important roles in iron homeostasis, the imbalance of which is a hallmark of FRDA.⁶⁹ Iron taken up by endocytosis eventually reaches lysosomes as the terminal compartment of this pathway, from where it is transferred to the mitochondria or cytosol. This process relies on proper lysosomal acidification by the V-ATPase,⁷⁰ as impairment of V-ATPase activity was shown to result in cellular iron deficiency, which eventually leads to mitochondrial dysfunction.⁷¹ Free iron in cells can lead to oxidative stress; therefore, it is often coupled to other molecules, such as polyamines in lysosomes, which can be internalized by membrane HSPGs. In this context, a study revealed that GPC1-HS modifications play essential roles in the uptake of spermine (a polyamine) via the endosomal pathway,^50,72 which in turn may have an impact on lysosomal iron metabolism.

Notably, two other recessively inherited ataxia types are caused by homozygous variants in NHE1 (MIM *107310) or CHP1 (MIM *606988),^73,74 which encode two proteins that interact with each other and are involved in ion regulation and pH homeostasis in mammalian cells. These functions are on a similar line to the role of the V-ATPase complex in pH regulation and vesicular trafficking. Therefore, the regulation of cellular pH homeostasis represents an intriguing aspect of the GPC1 interactome and merits further investigation in the future.

While the differences observed for the GPC1/GPC1 R142W interactomes in HEK293T cells provide insights into potentially mutation-dependent changes in cellular mechanisms, this model system naturally cannot fully reflect the situation in potentially disease-causing cells in situ. Therefore, additional studies are needed to further delineate possible disease pathomechanisms and explain the observed patient phenotype. This could include specialized cell types that are critical to the observed phenotype, such as neuronal cultures, or animal models that would allow for a more holistic view. For such endeavors, the GPC1 interaction partners identified in this study could provide a starting point for future studies.

Conclusions

We present a proteomic analysis of the GPC1 interactome that provides new insights into the cellular functions and localization of GPC1 and possible explanations for its involvement in neuronal function. Moreover, we suggest that the detected mutation in GPC1 might act as a disease modifier and worsen the FRDA phenotype. It is conceivable that a similar effect also applies to other forms of ataxia.

Data Availability Statement

The mass spectrometry proteomics data have been deposited in the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via PRIDE partner repository, with the dataset identifier PXD040023. The NGS genome data of the family are stored in the EGA database under access numbers EGAR00001408420 and EGAN00001366913. Uncropped images of the Western blots can be found under File S1.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jproteome.3c00402.

• (Figure S1) Supplement to Figure 1, confirmation of GPC1 antibody specificity; (Figure S2) supplement to Figure 2, HA-tagged GPC1 construct utilized for interactome studies; (Figure S3) supplement to Figure 3, unbiased analysis of the proteomic GPC1 interactome dataset; (Figure S4) supplement to Figure 3, Pearson correlation analyses and scatter plots for the quantitative DIA-MS data used in this study; (Figure S5) supplement to Figure 5, proteomic analysis of the impact of R142W on the GPC1 interactome (PDF)

• (Table S1) Supplement to Figure 1, homozygous variant list (XLSX)

• (Table S2) Spectronaut output for proteins (XLS)

• (Table S3) Spectronaut output for peptides (XLS)

• (Table S4) Supplement to Figure 2, raw protein list with log₂ adjusted measurement values (XLS)

• (Table S5) Supplement to Figure 3, GPC1 WT interactome (XLS)

• (Table S6) Supplement to Figure 3, GPC1 WT interactome GO analyses (XLS)

• (Table S7) Supplement to Figure 5, GPC1 R142W vs GPC1 WT differential interactome analysis (XLS)

• (Video S1) Supplement to Figure 1, index patient at 8 years of age showing wide-based ataxic unsteady gait aided by the parents (MP4)

• (File S1) Uncropped Western blots (PDF)

Author Present Address

^# Bijvoet Center for Biomolecular Research and Utrecht Institute for Pharmaceutical Sciences, Utrecht University, Utrecht, the Netherlands (P.M.)

Author Contributions

N.C.W., M.K., D.W., and B.W. performed conceptualization, design of experiments, and review and editing of the manuscript; N.C.W., M.K., P.M., and IB performed experiments; B.A. and G.H. clinically investigated the patient and the family members. M.K. performed the bioinformatic WGS analysis. N.C.W. made the figures; N.C.W., P.M., and B.W. performed the original draft preparation; all authors have read, corrected, and agreed to the published version of the manuscript.

This work was supported by the German Research Foundation Wi 945/19-1 (project ID 417989143), SFB1451 (project ID 431549029–A01), the European Community’s Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 2012-305121 “Integrated European–omics research project for diagnosis and therapy in rare neuromuscular and neurodegenerative diseases (NEUROMICS)”, and the Center for Molecular Medicine Cologne (project no. C18).

The authors declare no competing financial interest.

Notes

This study protocol was approved by the institutional review board and the ethics committee of the University of Cologne (approval number 13-022).

Notes

Written informed consent was obtained from legal guardians during recruitment using standardized forms.