Introduction

Glycosylphosphatidylinositol (GPI)-anchored proteins are a diverse set of plasma membrane-linked cell surface proteins [1]. The Biosynthesis of GPI-anchored proteins consists of the synthesis of GPI from phosphatidyl inositol and the subsequent attachment of GPI to proteins, both taking place in the endoplasmic reticulum (ER) membrane. Precursor proteins that are to be GPI anchored have a signal peptide for GPI attachment at the carboxyl-terminal. GPI transamidase in the ER cleaves the signal peptide and transfers the newly exposed carboxyl terminal of the protein to a GPI, thereby completing the synthesis of GPI-anchored proteins. GPI-anchored proteins are then transported to the cell surface through the secretory pathway. GPI transamidase is an enzyme complex consisting of five subunits called PIGK, PIGS, PIGT, PIGU, and GPAA1. PIGK is a single transmembrane protein that catalyzes the cleavage of the signal peptide [2, 3]. The other four proteins are likely to be involved in the recognition of the GPI signal peptide of precursor proteins [1]. Loss of either subunit of GPI transamidase impairs GPI transamidase activity and results in the elimination of GPI-anchored protein from the cell surface.

Pathogenic variants of GPI transamidase subunit genes in humans cause autosomal recessive neurological disorders. Variants in PIGK cause neurodevelopmental disorders with hypotonia and cerebellar atrophy, with or without seizures (OMIM 605087) [4, 5]. Similarly, PIGS pathogenic variants cause developmental and epileptic encephalopathy 95 (OMIM 610271) [6,7,8]. Variants in PIGT cause multiple congenital anomalies-hypotonia-seizures syndrome 3 (OMIM 610272) [9,10,11,12,13,14]. Variants in PIGU cause neurodevelopmental disorders with brain anomalies, seizures, and scoliosis (OMIM 608528) [15]. Finally, GPAA1 pathogenic variants cause GPI biosynthesis defect 15 (OMIM 603048), which is characterized by developmental delay, epilepsy, cerebellar atrophy, and osteopenia [16]. These neurological disorders mostly share clinical features such as developmental delay, hypotonia, cerebellar atrophy, seizures, intellectual disability, and characteristic faces, and are currently categorized as inherited GPI deficiencies [17]. However, it remains largely unclear how the loss of GPI-anchored proteins causes neurodevelopmental disorders.

In this study, we report a Japanese girl with neurodevelopmental disorder, cerebellar atrophy, seizures, hearing loss, growth impairment, dysmorphic facial features, and brachydactyly, who carried a novel homozygous variant of PIGK, NM_005482:c.481A > G,p. (Met161Val) caused by paternal isodisomy. Our rescue experiments using PIGK-deficient CHO cells revealed that this amino acid substitution affected GPI transamidase activity in a concentration-dependent manner. We also demonstrated that axonal localization of voltage-gated sodium channels, and thus generation of action potentials, were affected in pigk-deficient zebrafish neurons.

Materials and methods

Clinical descriptions

The female proband (III:4) was the second child of her healthy, nonconsanguineous Japanese parents (II:1 and II:5) (Fig. 1A). The first child was an early miscarriage of unknown cause. The third child was a girl healthy at 2 years and 3 months. There was no other family history to report. The patient was born at 40 weeks and 4 days of gestation. Her weight at birth was 3112 g (+0.1 SD), length was 49.0 cm (−0.4 SD), and head circumstance was 36.5 cm (+2.4 SD) (Supplementary Table S1). Although she showed no abnormality at birth, she was followed up due to a failure to thrive that became noticeable as generalized hypotonia and developmental delays (Table 1). Brain magnetic resonance imaging (MRI) showed cerebellar atrophy, decreased white matter, thin corpus callosum, wide opening of the Sylvian fissure, and larger ventricles in the brain at 1 year and 6 months of age (Fig. 1B). Although she passed the newborn hearing screening test, she was found to have bilateral moderate hearing loss at 1 year and 4 months. Treatment with tympanostomy tubes for otitis media was not effective for her hearing impairment, therefore she began wearing her hearing aids at age 2. A computed tomography (CT) scan revealed findings that were suspicious of bilateral ossicular malformations. Her growth impairment still existed at 4 years and 1 month: her weight was 12.1 kg (−2.1 SD), height was 92.8 cm (−1.9 SD) and head circumstance was 47.8 cm (−1.1 SD). She started walking at 2 years and 7 months and speaking at 3 years and 2 months. Her developmental quotient using the Kyoto Scale of Psychological Development 2020 was 50 at the age of 3 years and 6 months. She spoke only a few words but willingly tried to use nonverbal communication at the age of 4 years and 3 months. She began treatment with antiepileptic drugs at the age of 4 years to control recurrent seizures with a high fever. Neither ataxia, spasticity, nor involuntary movements had been presented during the follow-up period. These symptoms are not contradictory to those reported as neurodevelopmental disorders with hypotonia and cerebellar atrophy, with or without seizures, which are caused by genetic PIGK deficiencies [4, 5].

Fig. 1
figure 1

Clinical features and genetic analysis. A Pedigree of the family. The proband girl (III:4) was the second child of nonconsanguineous parents (II:1 and II:5). B Brain MRI of the patient at 1 year 6 months old. Cerebellar atrophy (asterisk), decreased white matter, thin corpus callosum (arrowhead), wide opening of the Sylvian fissures (arrows), and larger ventricles in brain were observed. C Integrative Genomics Viewer Images. The amino acid residues of PIGK are shown at the top and read counts are at left side of coverage track. Position of the candidate variant was indicated with black triangle. The c.481A > G missense variant of PIGK was homozygous in the proband (III:4) and heterozygous in the father (II:1). The mother had reference homozygous at c.481 (II:5). The coverage at the variant was comparative (proband: 50; father: 70; mother: 55). D Electropherogram of Sanger sequencing analysis. Heterozygous c.481A > G in the father (II:1); reference nucleotide c.481A in the mother (II:5); Homozygous c.481A > G in the proband (III:4). E Zygosity mapping using the WES data. For homozygous sequences of patients which passed data quality, the origin was estimated from the sequence data of the parents. The position of nucleotides estimated to have been inherited from the father was plotted as 1, the position of nucleotides estimated to have been inherited from the mother as -1, and the position of nucleotides that could still be inherited from either parent as 0. The horizontal axis indicates chromosome position

Full size image
Table 1 Clinical features and PIGK variants in the present case and reported cases in literatures
Full size table

Whole-exome sequencing analysis

After obtaining informed consent from the patient’s parents, genomic DNAs were extracted from the peripheral blood in the family and trio-based whole-exome sequencing (WES) and filtering analysis were performed as previously described [18]. In brief, the sequence library was prepared using a Human All Exon V6 Kit (Agilent Technology, Santa Clara, CA, USA) and sequenced using NovaSeq (Illumina, CA, USA) with 150-bp paired-end reads. Sequence reads were aligned to GRCh38 and annotated using CompStor NOVOS and CompStor Insight (OmniTier, CA, USA). First, high-quality variants with allele frequencies greater than 0.01 in gnomAD, 14KJPN (jMORP), and our in-house exome variant data were removed. Next, the variants were narrowed down based on the assumed modes of inheritance, such as autosomal dominant, autosomal recessive, and X-linked inheritance. After filtration following this procedure, four variants, three homozygous variants, and one de novo variant were segregated (Supplementary Table S2). We further curated these variants by searching OMIM and PubMed databases, and prediction tools, CADD, Polyphen-2, SIFT, and MutationTaster. The nucleotide sequence of the variant region was confirmed by Sanger sequencing (ABI3130, Life Technologies, CA, USA). Genomic DNA was amplified using Ex-Taq (TAKARA, Shiga, Japan) with the forward primer; 5′-TAACTGGGAGGATCCCACCTAGT-3′ and reverse primer; 5′-CAGTTGGTAATCTTCATGTACAGC-3′. Electropherograms of the sequencing result were obtained using reverse primer. Zygosity analysis on chromosome 1 using the WES data was performed previously reported using an in-house program [19]. In brief, homozygous nucleotides of the patients were selected and their origin was estimated by reference to the parental genotypes. Nucleotide positions that could have been inherited from the father were plotted as 1, those that could have been inherited from the mother as −1, and those that could have been inherited from either parent as 0.

In silico three-dimensional structural analysis of the PIGK protein

Data on the cryo-EM structure of the human GPI transamidase complex (PDB code: 7WLD) [20, 21] were imported into the Molecular Operating Environment (MOE). The methionine at position 161 was replaced with valine or leucine, and the three-dimensional views around the amino acids were calculated.

Phylogenetic analysis

Amino acid sequences of vertebrate PIGK proteins were obtained from the NCBI database. The accession numbers are as follows; Human: NP_005473; Rhesus monkey: NP_001247873; Mouse: NP_079938; Chicken: NP_001026449; Common lizard: XP_034977403; Frog (Xenopus tropicalis): NP_001072709; and Zebrafish: NP_001002149. Genetyx-Mac Ver. 22 (Genetyx Corporation, Tokyo, Japan) was used to construct a protein alignment.

CHO cell analysis

The empty pME or pTK vectors or human PIGK (wild-type or p.Met161Val) expression plasmids were electroporated to 107 cells of PIGK-deficient Chinese hamster ovary (CHO) cells (clone 10.2.2) [22]. Two days after electroporation, the 2 × 104 cells were analyzed in a single run by fluorescence-activated cell sorting (FACS) using a flow cytometer (MACSQuant Analyzer; Miltenyi Biotec, Bergisch Gladbach, Germany) with FlowJo software (Tommy Digital, Tokyo, Japan) as described previously [5]. The following antibodies were used for cell labeling: mouse anti-human CD59 (clone 5H8); anti-human decay accelerating factor (DAF) (clone IA10, BD Biosciences); mouse anti-hamster urokinase-type plasminogen activator receptor (uPAR) (clone 5D6); and phycoerythrin-conjugated anti-mouse IgG antibody. Whole-cell lysates of 106 CHO cells were subjected to SDS-PAGE and Western blotting. Expression of PIGK proteins was detected by goat anti-GST polyclonal antibody (Sigma-Aldrich, St. Louis, USA), followed by HRP-conjugated anti-goat IgG. For the loading control, GAPDH was detected by mouse anti-GAPDH (AM4300, Invitrogen, Waltham, MA, USA), followed by HRP-conjugated anti-mouse IgG.

Animals

Zebrafish (Danio rerio) were reared and maintained in 1.7 L tanks in a recirculating system (Meito System, Nagoya, Japan) at 28.5 °C under a 14 h light and 10 h dark photoperiod. Larvae were fed paramecia and Gemma Micro ZF 75 (Skretting, Westbrook, ME, USA) twice daily from 5 to 30 days postfertilization (dpf). Juvenile fish were fed brine shrimp (Tokai Guppy, Okazaki, Japan) and Gemma Micro ZF 75 twice daily from 30 to 90 dpf. Adult fish were fed brine shrimp and Otohime B2 (Marubeni Nissin Feed, Tokyo, Japan) twice daily after 90 dpf. We obtained a zebrafish pigktt261 mutant allele from the Zebrafish International Resource Center. This pigktt261 allele carries a T to C mutation in the first exon of the pigk gene, disrupting the first methionine codon [23]. The Tg(zCREST2-hsp70:GFP) transgenic fish [24] used to visualize sensory Rohon–Beard (RB) neurons in electrophysiology were provided through the National BioResource Project of Japan.

Zebrafish analyses

For rescue experiments, full-length cDNAs encoding human PIGK were cloned into the expression vector pCS2+ using the In-Fusion HD Cloning Kit (Clontech, Mountain View, CA, USA) according to the manufacturer’s manual. The p.Met161Val, p.Gln33TER, p.Ser53Phe, p.Ala87Val, p.Tyr160Ser, p.Asp204His, and p.Cys275Arg mutations were introduced into the human PIGK expression construct by site-directed mutagenesis, as described previously [25]. Primers used for cloning and mutagenesis are listed in Supplementary Table S3. Expression constructs were used to generate human PIGKWT, PIGKM161V, PIGKQ33X, PIGKS53F, PIGKA87V, PIGKY160S, PIGKD204H, and PIGKC275R mRNAs using the mMESSAGE mMACHINE SP6 Kit (Thermo Fisher Scientific, Waltham, MA, USA) as described previously [25]. Human PIGK mRNAs (100 pg) were injected into 1-cell-stage zebrafish embryos produced by crossing PIGKtt261 heterozygous fish. For genotyping, the genome was extracted from mRNA-injected embryos, as described previously [26]. The genomic pigk region surrounding the T to C mutation (ATG to ACG in the first methionine codon) was amplified by PCR and subjected to direct sequencing using an Applied Biosystems 3500 Genetic Analyzer (Thermo Fisher Scientific).

For the touch response, tactile stimulation was applied to the tails of zebrafish embryos at 48 h postfertilization (hpf). Touch responses were recorded at 200 frames per second using a high-speed camera HAS-220 (Ditect, Tokyo, Japan), as described previously [27]. Embryos that showed swimming following a touch were judged as normal, whereas embryos that did not respond to a touch were judged as unresponsive. Mutant embryos that showed the recovery of touch response displayed normal swimming locomotion in appearance.

For in situ hybridization, full-length cDNAs encoding zebrafish pigk were cloned into the expression vector pCS2+. Primers used for cloning and mutagenesis are listed in Supplementary Table S3. DIG-labeled riboprobes were synthesized and used for RNA labeling with NBT/BCIP (Roche, Basel, Switzerland) according to established procedures [28].

For electrophysiology, zebrafish embryos (48 hpf) carrying zCREST2-hsp70:GFP transgene [24] were used to visualize RB cells. The dissection protocols for in vivo patch recordings in RB neurons have been described previously [23, 29]. The resting membrane potentials of the patched RB cells (−58 ~ −68 mV) were comparable between wild-type and pigk mutant embryos.

For immunohistochemistry, skin-removed zebrafish embryos (48 hpf) were fixed in 4% paraformaldehyde and subjected to immunolabeling, as described previously [28]. The following primary antibodies were used: rabbit anti-voltage-gated sodium channel Pan (SP19, 1:500; Sigma-Aldrich) and mouse monoclonal anti-HuC/HuD (16A11, 1:500; Thermo Fisher Scientific). Alexa 488-conjugated anti-rabbit IgG and Alexa 568-conjugated anti-mouse IgG (1:500; Thermo Fisher Scientific) were used as the secondary antibodies. Fluorescence images were captured using a confocal microscope TCS SP5 (Leica Microsystems, Wetzlar, Germany). The Nav labeling at the cell body was observed in all wild-type and pigk mutant RB cells, but Nav labeling at axons was seen in 20% of wild-type RB cells but not at all in mutant RB cells.

Results

Genetic analysis

We identified a novel homozygous variant of the PIGK gene, NM_005482:c.481A > G,p. (Met161Val), on chromosome 1 in the patient. The father was a heterozygous carrier of the variant but the mother was not (Fig. 1C, D). Since all patient variants detected in the exon and splice regions of chromosome 1 were homozygous and no copy number changes were detected, the origin of the homozygous variants on chromosome 1 was estimated. The results showed that these homozygous variants were inherited exclusively from the father, suggesting a paternal disomy (Fig. 1E).

In silico three-dimensional structural prediction of the PIGK protein

The missense variant of PIGK resulted in an amino acid substitution from methionine to valine at position 161 in the third β-sheet (β3) (Fig. 2A). The methionine residue in the β3 was conserved among mammals, birds, lizards, and amphibians, but not in zebrafish, in which methionine was replaced with leucine (Supplementary Fig. S1). We calculated the three-dimensional structure of the PIGK protein using MOE software with the cryo-EM structures of human GPI transamidase [20, 21]. In wild-type human PIGK protein, Met161 on the β3 was predicted to exist at a certain distance from Ile203 on the β4, which ran parallel to the β3, and at a close distance from Phe131 on the second alpha helix 2 (α2) (Fig. 2B; Supplementary Fig. S2A, B). In the Met161Val substitution observed in patients, the Val161 residue was presumed to be located very close to Ile203 on the β4 but with a greater distance to Phe131 on the α2 (Fig. 2C; Supplementary Fig. S2C, D). In the Met161Leu substitution as in the zebrafish case, the Leu161 residue was predicted to keep a certain distance from Ile203 (Fig. 2D; Supplementary Fig. S2E, F). The distance from Leu161 to Phe131 was predicted to be comparable with that from Met161 to Phe131 as in the wild-type human PIGK protein (Fig. 2D; Supplementary Fig. S2E, F).

Fig. 2
figure 2

In silico three-dimensional analysis of PIGK protein. A Domains and variants of PIGK protein. Arrows above the protein illustration indicate the position of amino acid residues critical for PIGK function. Arrows below the protein illustration indicate the position of amino acid residues affected in pathogenic variants. Orange and pink boxes indicate regions of the α helix and β strands, respectively. Note that the p.Met161Val variant (red) found in this study is located close to catalytic active sites of cysteine protease function. SP signal peptide, TM transmembrane domain. B Ribbon diagram of the Met161 region in the PIGK protein. C Ribbon diagram of the p.Met161Val substitution in the PIGK protein. D Ribbon diagram of the Met161Leu substitution in the PIGK protein. Note that the Met161 position is located in the β3 strand that runs parallel to the β4

Full size image

Functional analysis of p.Met161Val variant in CHO cells

The PIGK protein is essential for the GPI transamidase activity, which enables cell surface expression of GPI-anchored proteins. To address whether the Met161Val substitution of PIGK affected GPI transamidase activity, we employed CHO cells for in vitro rescue assay. The PIGK-deficient CHO cells were electroporated with an empty pME expression vector or pME vector carrying the human wild-type or variant PIGK gene. GPI-anchored proteins, which came out to the cell surface by rescue expression of functional PIGK in PIGK-deficient CHO cells, were labeled with antibodies against GPI-anchored proteins and investigated using FACS. Expression of wild-type or Met161Val variant PIGK rescued cell surface expression of GPI-anchored proteins such as CD59, DAF, and uPAR in PIGK-deficient CHO cells (Fig. 3A). As the pME vector drives gene expression under a strong SRα promoter, we next used a pTK vector that drives gene expression under a weak thymidine kinase promoter. Expression of wild-type PIGK using pTK expression vector successfully rescued the cell surface expression of GPI-anchored proteins, as in the use of the pME vector. However, the expression of the Met161Val PIGK variant using the pTK vector restored the expression of GPI-anchored proteins less efficiently than that of wild-type PIGK (Fig. 3B). Our Western blots confirmed that both wild-type and variant PIGK proteins were expressed at similar levels in these cells (Fig. 3C). These results indicate that the Met161Val substitution of PIGK reduced GPI transamidase activity and thus diminished cell surface expression of GPI-anchored proteins.

Fig. 3
figure 3

In vitro functional analysis of PIGK p.Met161Val variant in PIGK-deficient CHO cells. A Wild-type or M161V variant PIGK was expressed under the strong SRα promoter (pME vector) in PIGK-deficient CHO cells. Cell surface expression of GPI-anchored proteins (CD59, DAF, and uPAR) was evaluated by FACS. B Wild-type or M161V variant PIGK was expressed under the weak promoter (pTK vector) in PIGK-deficient CHO cells. The right-sided peaks indicate higher levels of GPI-anchored proteins at the cell surface. Note that the surface expression of GPI-anchored proteins was diminished in M161V variant expression (blue line), but not in wild-type PIGK expression (red line), only when the pTK vector was used. Light blue line: electroporation of empty vectors; dark gray peaks: wild-type CHO cells (3B2A); light gray peaks: isotype control. C Western blots of PIGK proteins. Both wild-type and variant (M161V) PIGK proteins were expressed in PIGK-deficient CHO cells. GAPDH is a loading control

Full size image

Functional analysis of p.Met161Val and the other variants in zebrafish

To further ascertain whether the Met161Val substitution and the other previously reported variants hindered PIGK function in vivo, we employed pigk mutant zebrafish as a PIGK-deficient animal. Our group and another group have reported that pigk-deficient zebrafish failed to show tactile-evoked escape swimming due to the reduction of voltage-gated sodium currents in sensory neurons at 48 hpf [23, 27]. We obtained pigk mutant zebrafish embryos by crossing pigk heterozygous fish, injected human PIGK mRNA into the embryos at the 1-cell stage, and assayed the tactile response at 48 hpf followed by genotyping (Fig. 4A). Injection of wild-type human PIGK mRNA rescued tactile-induced normal swimming in 75% of mutant embryos (Fig. 4B), while that of the Met161Val variant mRNA restored the touch response only in 15% of mutants (Fig. 4C, D), suggesting that the Met161Val substitution reduced the function of PIGK. Similarly, the injection of PIGK mRNA carrying disease-relevant missense variants (Ser53Phe, Ala87Val, Tyr160Ser, Asp204His, and Cys275Arg) showed a compromised recovery in tactile-evoked swimming. However, all of the mutant embryos injected with PIGK mRNA carrying a Gln33Ter nonsense variant, as well as uninjected mutant embryos, failed to show the restoration of the touch response. These results indicate that the Gln33Ter variant appeared to be a loss-of-function variant, whereas missense variants such as Met161Val variant appeared to retain a compromised PIGK function.

Fig. 4
figure 4

In vitro functional analysis of PIGK variants in PIGK-deficient zebrafish embryos. A Schematic illustration of the functional analysis using zebrafish. Zebrafish embryos obtained by crossing pigk (+/−) fish were injected with human PIGK mRNA (wild-type or variant) and subjected to a tactile response followed by genotyping of the pigk gene. B Injection of human wild-type PIGK mRNA in pigk mutants restored touch-evoked swimming. C Injection of human M161V PIGK mRNA in pigk mutants failed to restore the tactile response. D Histogram of the rescue of the tactile response. Statics were obtained using the chi-squared test with Yates’ correction. Note that the injection of human Q33X PIGK mRNA did not cause recovery of the tactile response, whereas injection of the other variant mRNA caused weak recovery. The P values of the chi-squared test were as follows; M161V: 4.7E-04; Q33X: 4.3E-05; S53F: 2.5E-05; A87V: 4.4E-03; Y160S: 2.3E-03; D204H: 8.3E-05; C275R: 7.1.E-04

Full size image

The defect of PIGK affects the subcellular localization of voltage-gated sodium channels in vivo

To address neuronal deficits in pigk mutant zebrafish, we focused on sensory RB neurons, which are necessary for mechanoreception and, thus, for tactile-evoked escape response in zebrafish embryos. Our in situ hybridization confirmed that pigk is indeed expressed in RB neurons, which are located at the most dorsal region in the spinal cord (Fig. 5A, B). We also performed in vivo patch-clamp recordings of RB neurons and confirmed that current injections elicited action potentials in wild-type RB cells, but not in mutant RB cells, recapitulating that voltage-gated sodium channels are affected in pigk-deficient RB neurons (Fig. 5C, D). To further investigate the defects of voltage-gated sodium channels, we assessed subcellular localization of voltage-gated sodium channels. Anti-Pan Nav immunolabeling revealed the distribution of voltage-gated sodium channels at cell bodies and axons in wild-type RB neurons (Fig. 5E–G). On the other hand, labeling of voltage-gated sodium channels was observed at cell bodies but not at axons in pigk mutant RB cells (Fig. 5H–J). These results indicated that PIGK is necessary for the proper localization of voltage-gated sodium channels in zebrafish neurons and suggested that GPI-anchored proteins play a role in the transport of voltage-gated sodium channels. This may partly account for the neurological symptoms of patients harboring PIGK variants (see Discussion).

Fig. 5
figure 5

Localization and function of voltage-gated sodium channel is impaired in pigk-deficient zebrafish neurons. A, B In zebrafish, pigk is expressed in Rohon–Beard sensory neurons at 48 hpf. C Patch-clamp-mediated current injection generated action potentials in wild-type Rohon–Beard neurons. D Current injections failed to elicit an action potential in pigk-deficient Rohon–Beard cells. E–G The Nav proteins were observed at soma and axons in wild-type Rohon–Beard neurons. H–J The Nav proteins were observed at soma, but not at axons in pigk-deficient Rohon–Beard cells. HuC is a marker of the cell body in Rohon–Beard neurons

Full size image

Discussion

In this study, we reported a Japanese girl with a neurodevelopmental delay, hypotonia, cerebellar atrophy, febrile seizures, hearing loss, growth impairment, dysmorphic facial features, and brachydactyly, who had the novel homozygous variant of PIGK, c.481A > G,p.(Met161Val) inherited by paternal isodisomy. The variant presented a very low allele frequency in the population of 0.000013 for gnomAD and 0.00007 for 14KJPN, while no homozygote was registered in those databases. The protein effect prediction program predicted this as deleterious and possibly damaging for SIFT and PolyPhen-2, respectively, with a CADD score of 23.1. The clinical manifestations of the patient were mostly similar to those of previously reported patients harboring PIGK pathogenic variants. Furthermore, we performed functional analysis and revealed that the variant affected PIGK functions. Therefore, the variant was classified as pathogenic according to the ACMG/AMP guidelines (PS3, PM2, PP2, PP4).

Structural prediction of the p.Met161Val PIGK protein

The Val161 was predicted to be located on a parallel β-sheet. Indeed, most of the pathogenic variants reported in the literature were located on parallel β-sheet structure (L86P, Y160S, D204H) or proximity to β-sheet structure (S53F, A87V, D87N, M246K; Supplementary Fig. S3A, B) [5]. Since the amino acid substitution, Met161Val changes the direction of the side chain from Phe131 to Ile203, it might distort the parallel β-structure. In addition, Ile203 is only three amino acids away from the Cys206 [20, 21], which is the catalytic residue for cysteine protease function, resulting in likely to affect protease function. Conversely, the Leu161 residue in zebrafish PIGK protein appeared to maintain the distance between Leu161 and Ile203 without distorting the parallel β-structure. These structural predictions account for the finding that the Met161Val substitution compromised GPI transamidase function in CHO cells.

Pathogenic variants of the PIGK gene and their function

To date, 14 patients in 10 independent families with the neurodevelopmental disorders with hypotonia and cerebellar atrophy, with or without seizures (OMIM 605087), have been reported. These patients had homozygous or compound heterozygous pathogenic variants in PIGK [4, 5]. The current study is the first report of a Japanese patient with PIGK pathogenic variant. Among the total 15 cases, 10 missense variants of PIGK (Ser53Phe, Leu86Pro, Ala87Val, Asp88Asn, Tyr160Ser, Met161Val, Ala184Val, Asp204His Met246Lys, and Cys275Arg) have been identified. The other pathogenic variants of PIGK were one nonsense (Gln33Ter), one frame-shift (Ile30Tyrfs*10), and one splicing junction site (between Glu31 and Asp32). These latter three variants may disrupt the PIGK protein due to protein truncation or nonsense-mediated mRNA decay. Thus, these three variants are presumably the complete loss of function. These variants were identified as compound heterozygous variants with missense variants, but not as homozygous variants, suggesting that complete loss of PIGK function in humans is not viable and that the missense variants reported in patients may not be a complete loss of PIGK function. In fact, our in vitro rescue experiments showed that cell surface expression of GPI-anchored proteins in PIGK-deficient CHO cells was only partially recovered by low-level expression of the PIGK Met161Val variant and completely recovered by high-level expression of the same variant, indicating that the Met161Val variant was a hypofunctional variant, just like a previously characterized pathogenic Cys275Arg variant [5]. Our in vivo rescue assay using zebrafish also revealed that expression of the nonsense PIGK variant (Gln33Ter) failed to recover the touch response in pigk-deficient embryos, whereas that of missense variants (Ser53Phe, Ala87Val, Tyr160Ser, Met161Val, Asp204His, and Cys275Arg) recovered tactile-evoked swimming but at lower rates. Hence, missense variants of PIGK may retain compromised PIGK function and are not complete loss of function.

PIGK is necessary for the proper localization of voltage-gated sodium channels

We found that subcellular localization of voltage-gated sodium channels is affected in pigk-deficient zebrafish neurons. As febrile seizures, which were present in the current patient, are caused by genetic variants of voltage-gated sodium channels [30, 31], the involvement of PIGK in the regulation of voltage-gated sodium channels may partly account for the pathological symptoms of the current patient. Voltage-gated sodium channels consist of 24 transmembrane segments and thus are membrane-spanning proteins, but not GPI-anchored proteins. One would wonder how a deficiency of GPI-anchored proteins affects the localization of voltage-gated sodium channels. We assume that specific GPI-anchored proteins interact with voltage-gated sodium channels that, in turn, promote the intracellular transport of voltage-gated sodium channels to the axonal membrane. However, GPI-anchored proteins that interact with voltage-gated sodium channels have not been identified. Elucidation of such GPI-anchored proteins will push the understanding of the pathological symptoms of inherited GPI deficiencies.