The P429L loss of function mutation of the human glycine transporter 2 associated with hyperekplexia

Glycine transporter 2 (GlyT2) mutations across the entire sequence have been shown to represent the presynaptic component of the neurological disease hyperekplexia. Dominant, recessive and compound heterozygous mutations have been identified, most of them leading to impaired glycine uptake. Here, we identified a novel loss of function mutation of the GlyT2 resulting from an amino acid exchange of proline 429 to leucine in a family with both parents being heterozygous carriers. A homozygous child suffered from severe neuromotor deficits. We characterised the GlyT2P429L variant at the molecular, cellular and protein level. Functionality was determined by glycine uptake assays. Homology modelling revealed that the mutation localises to α‐helix 5, presumably disrupting the integrity of this α‐helix. GlyT2P429L shows protein trafficking through various intracellular compartments to the cellular surface. However, the protein expression at the whole cell level was significantly reduced. Although present at the cellular surface, GlyT2P429L demonstrated a loss of protein function. Coexpression of the mutant with the wild‐type protein, reflecting the situation in the parents, did not affect transporter function, thus explaining their non‐symptomatic phenotype. Nevertheless, when the mutant was expressed in excess compared with the wild‐type protein, glycine uptake was significantly reduced. Thus, these data demonstrate that the proline residue at position 429 is structurally important for the correct formation of α‐helix 5. The failure in functionality of the mutated GlyT2 is most probably due to structural changes localised in close proximity to the sodium‐binding site of the transporter.


| INTRODUCTION
Glycine transporter 2 (GlyT2) is a sodium/chloride-dependent glycine transporter mainly located in presynaptic membranes of spinal cord and brainstem interneurons (Chalphin & Saha, 2010). GlyT2 is crucial for the transport of glycine from the synaptic cleft back into the presynaptic compartment where the neurotransmitter glycine will be packaged by the vesicular inhibitory amino acid transporter (VIAAT) into synaptic vesicles for a second release (Eulenburg, Armsen, Betz, & Gomeza, 2005;Harvey, Topf, Harvey, & Rees, 2008).
Two classes of glycine transporters exist, GlyT1 and GlyT2, which are suitable therapeutic targets for the treatment of central and peripheral nervous system disorders (Harvey & Yee, 2013). The activities of both transporters depend on the concentrations of extracellular glycine, Na + and Cl − ions. In contrast to GlyT1 which has a 2 Na + /1 Cl − / glycine stoichiometry, GlyT2 exhibits a 3 Na + /1 Cl − /glycine stoichiometry Roux & Supplisson, 2000;Supplisson & Roux, 2002). The GlyT1 is mainly localised at astrocytes of the caudal CNS regulating the availability of glycine as a coagonist for the NMDA receptor (Zafra et al., 1995). In contrast, GlyT2 is restricted to inhibitory synapses in the spinal cord, brainstem and cerebellum (Chalphin & Saha, 2010;Muller, Le Corronc, Triller, & Legendre, 2006). The X-ray crystal structures of both of these transporters are not yet known. Homology modelling with the closely related eukaryotic dopamine transporters (DAT) or the prokaryotic leucine transporter (LeuT) has been used to predict structural consequences of GlyT2 mutations (Penmatsa, Wang, & Gouaux, 2013;Wang, Penmatsa, & Gouaux, 2015;Yamashita, Singh, Kawate, Jin, & Gouaux, 2005). The transmembrane domain (TMD) presumably consists of 12 α-helices, with short intra-or extracellular loops connecting these TM helices. The N-and C-termini are localised intracellularly. TM helices 1-5 and TM helices 6-10 compose two structural motifs in opposite direction within both transmembrane segments. TM helix 1 and TM helix 6 constitute a break point in the middle of the TMD and are composed of two segments each (TM1a, TM1b, TM6a, TM6b). These two TM helices together with TM3 and TM8 represent the inner ring of the transporter which accommodates the substrateand Na + -binding sites (Gomeza et al., 2003). The GlyT2 protein is transported as a glycosylated dimer or oligomer to the cellular membrane (Bartholomaus et al., 2008;Fernandez-Sanchez, Diez-Guerra, Cubelos, Gimenez, & Zafra, 2008;Yamashita et al., 2005).
Sequencing of genomic DNA from patients with startle disease identified missense, frameshift and nonsense mutations distributed over the entire 16 exons of the SLC6A5 gene (Rees et al., 2006). The majority of GlyT2 mutations in patients are compound heterozygous demonstrating single recessive mutations from the unaffected mother and father; however, when inherited together, they result in two affected SLC6A5 alleles associated with a disease phenotype. In our study, we characterise the GlyT2 mutation P429L originally identified in an infant with startle disease homozygous for the DNA polymorphism c.1286C > T. Both parents were carriers of this polymorphism and harboured a second polymorphism c.1387G > A also located in exon 8. This second polymorphism was not identified in the affected infant. The family consists of two additional healthy children with the identical genotype compared with the parents (in exon 8) and two affected children homozygous for c.1286 C > T but negative for the second polymorphism c.1387 G > A (one child died at the age of 15 months). The mutation c.1286 C > T resulted in an amino acid exchange of a proline at amino acid position 429 into a leucine at the protein level. Our molecular characterisation identified lower whole cell expression accompanied by a loss of glycine transport. Homology modelling revealed that the exchange of the proline to leucine destroys the kink in the α-helix 5 produced by the proline and may affect the mobility of the adjacent α-helix 8, which mediates binding of sodium ions. Consequently, P429L is considered a loss of function mutation of GlyT2.

| Allele sequencing
To analyse the allelic distribution of the genomic mutations from different family members, exon 8 was amplified by PCR from the gDNA, cloned into a plasmid, and various clones were sequenced (n = 10-12).

| Site-directed mutagenesis
For generation of the GlyT2 P429L mutant, mutagenesis primers were designed (Life Technologies, Darmstadt, Germany): forward mutated primer 5′GGCCACGTTCCTGTATGTCGTACT3′, reverse primer 5′GTACGACATACAGGAACGTGGCC3′. In addition, parental primers were designed including an additional restriction site to allow easy cloning (Cla I, Hind III): forward parental primer 5′TAG GTG ACA CTA TAG AAT AAC ATC3′, reverse parental primer 5′GTA ACC ATT ATA AGC TGC AAT AAA CAA GTT3′. The first PCR generated two amplimers which overlapped in the region of the mutagenesis primer. Using both initial fragments in a second PCR as templates, two cycles without additional primer were run to let the 3′OH end of both strands be filled by the polymerase. Then, the two parental primers were added and the full length mutated GlyT2 was amplified in 28 additional PCR cycles. The PCR reaction contained 100 ng plasmid DNA (human GlyT2), 10 pmol of each primer, 5× HotStar HiFidelity PCR buffer (20 μl, included dNTPs), 1 μl of HotStar HiFidelity DNA Polymerase (2.5 U/μl, Qiagen, Hilden, Germany) and adjusted to 100 μl with sterile water. The PCR conditions were 5 min denaturation at 95°C, 5 min annealing at 50°C and 5 min elongation at 72°C. The final PCR product was cut with the newly introduced restriction sites for Hind III and Cla I and ligated into the mycGlyT2/pRK5 plasmid. The final construct was verified for the mutation by sequencing (LGC Genomics, Berlin, Germany). Darmstadt, Germany). The medium was supplemented with glutaMAX (200 mM) and sodium pyruvate (100 mM), penicillin (10,000 U/ml)/streptomycin (10,000 μg/ml) and 10% foetal calf serum. Cells were grown at 37°C and 5% CO 2 . At a confluency of 75% (24 h after plating), cells were transfected using either the calcium phosphate precipitation method or lipofectamine. For calcium phosphate precipitation, cells were washed 6 h post-transfection and used either for immunocytochemical stainings and protein biochemical analysis 24-48 h after transfection. To perform glycine uptake assays, lipofectamin transfection of HEK293 cells was preferred.

| Cell lines
COS7 cells were used for the compartmental analysis due to their large cytoplasm. Cells were obtained from ATCC (CRL-1651; ATCC -Global Biosource Center, Manassas, VA, USA). COS7 cells were transfected using dextran and PBS (phosphate-buffered saline) combined with plasmid DNA. Following 30 min at 37°C and 5% CO 2 , cells were washed and incubated with chloroquine for an additional 2 h to increase transfection efficiency. After 2 h, chloroquine was washed away, and fresh medium was provided to the cells for another 48 h before cells were used for immunocytochemical stainings.

| Analysis of transfection capacities
GlyT2 wt and GlyT2 P429L were co-transfected with GFP plasmid and stained 48 h post-transfection. An Olympus IX81 microscope with Olympus FV1000 confocal laser scanning system was used to acquire fluorescence images. The FVD10 SPD spectral detector and diode lasers of 405 nm (DAPI), 495 nm (GFP) and 550 nm (Cy3) (Olympus, Tokyo, Japan) were used. Five images of three independent transfections (15 images total) and stainings were taken using Olympus UPLFLN 20× objective. Images were further processed, and cells were counted using ImageJ (1.52)/Fiji software (Schindelin et al., 2012). The number of DAPI-positive nuclei was used as total number of cells per image. The numbers of GFP and Cy3 positive cells (GlyT2 was stained with the monoclonal antibody anti-myc and the secondary goat anti-mouse Cy3 antibody) were calculated in percent dependent on the total amount of DAPI-positive cells. Mean values and SEM were calculated.

| Whole cell lysate and membrane preparation
Forty-eight hours post-transfection, whole cell lysates were obtained from transfected HEK293 cells using the protein extraction reagent CytoBuster (Novagen/Merck KGaA, Darmstadt, Germany). In brief, transfected cells from a 10 cm dish were washed once in PBS pH 7.4 and supplied with 500 μl CytoBuster reagent. Following 5 min incubation at 21°C, cells were collected by scraping them from the dish and centrifuged for 5 min at 4°C (16,000 g). Solubilised protein was located in the supernatant and stored at −80°C. For membrane protein analysis, crude cell membranes were prepared from cells transfected with either GlyT2 wt or GlyT2 P429L (Sontheimer et al., 1989).

| Endo H and PNGase F treatment
Endo H and PNGase F were used for deglycosylation experiments according to the manufacture's protocol (New England Biolabs, Ipswich, MA, USA).

| Glycine uptake assay
To perform the [³H]-glycine uptake assay, 96-well plates were gelatine coated. A total of 15,000 HEK293 cells were seeded per well and transfected 24 h after seeding with lipofectamine (Lipofectamine2000, Life Technologies, Darmstadt, Germany). Transfection with GFP served as MOCK control and to control of the transfection efficiency. Cells were either transfected with GlyT2 wildtype, GlyT2 P429L or defined ratios of 1:1 or 1:10 of both constructs (the total amount of plasmid was the same for all conditions used). For the uptake assay, sterile-filtered HBS (HEPES-buffered saline) buffer pH 7.4 (5 mM KCl, 1 mM MgSO 4 , HEPES-Tris, pH 7.4, 10 mM Glucose, 1 mM CaCl 2 , 150 mM NaCl,) was used. The transport medium contained a concentration series of glycine. To reach final concentrations of glycine in a series of 0, 25, 50, 100, 250, 500, 750, 1000 µM, the [ 3 H] glycine (stock 20.9 μM; 47.8 Ci/mmol, Perkin Elmer, Waltham, MA, USA) was used in a concentration series of 25, 50, 100, 250, 500, 750 and 1000 nM and supplied with 10 mM cold glycine to reach the final glycine concentrations (example: 1000 nM = 1 μM [ 3 H] glycine was supplemented with 999 μM cold glycine to reach 1000 μM = 1 mM glycine in total) (Fornes, Nunez, Aragon, & Lopez-Corcuera, 2004;Gimenez et al., 2012). Cells were washed three times in HBS buffer and supplied with 80 μl of the appropriate transport medium with different glycine concentrations for 10 min at 37°C. All samples were done in triplicates. Cells were washed again three times with HBS buffer. Cell lysis was performed with 140 μl cold sterile H 2 O for 10 min on ice. Lysed cells were homogenised. 100 μl were used for counting the amounts of glycine uptake. Counting was performed using the Perkin Elmer scintillation counter (Tri-Carb 2910TR, Perkin Elmer, Rodgau, Germany). 20 μl of lysate was used to determine the protein concentration in each probe, required for the calculation of the final uptake rate. Protein quantification was done using the Bradford assay (Bio Rad, Munich, Germany). The GFP-transfected cells refer to the endogenous transporter activity in HEK293 cells. There was no background correction done by subtracting these values from the GlyT2 transporter activities.

| Statistical data analysis
Western blots from whole cell protein fractions were quantified using the Image J software (NIH). GlyT2 protein bands were normalized to β-actin (housekeeping protein). Microcal Origin (Microcal Software, Inc.) was used to analyse the glycine uptake capability. The equation y = v max * x n /(K n + x n ) was used to determine the maximal velocity (v max ) of the uptake reaction as well as the half-maximal velocity referred to as K m value. The graphs were generated using the Origin 6.0 software. Statistical data analysis was performed using Student's t test (analysis of variance) or one-way ANOVA. Significance values are represented with *p ≤ 0.05, **p ≤ 0.01 and ***p ≤ 0.001.

| Homology modelling and sequence alignment
The GlyT2 homology model was built with SWISS-MODEL (Arnold, Bordoli, Kopp, & Schwede, 2006) by using the Drosophila melanogaster Dopamine transporter (DAT) (PDB:4XP4) (Wang et al., 2015) as the template model, which shares a sequence identity of ~49%. In silico site-directed mutagenesis was performed in COOT (Emsley & Cowtan, 2004) by taking into consideration clashes with surrounding residues while retaining proper stereochemistry of the mutated residues. Figures involving structural representation were prepared in PyMol (www.pymol.org). The multiple sequence alignment was performed by using Clustal omega (Sievers et al., 2011), and the representation was generated through the ESPript server (Robert & Gouet, 2014).

| Patient description
The patient was the fourth child of two consanguineous parents, both were carriers of two polymorphisms in the coding region of the SLC6A5 gene. The polymorphisms were located in exon 8 of the SLC6A5 gene. Both parents (I/1 and I/2) were heterozygous for the determined polymorphisms c.1286 C > T and c.1387 G > A (Fig. 1a-c). At the protein level, the polymorphisms result in missense mutations of the encoded GlyT2 protein, P429L and D463N.

F I G U R E 1
Overview about family pedigree and sequence analysis. (a) The family pedigree shows the two heterozygous parental carriers (generation I; I/1 and I/2) and their four children (generation II; II/1-4). Note, the genotype of child 1 (II/1, grey square) is not known due to a lack of material for sequencing analysis. (b,c) Sequence analysis of exon 8 amplified from genomic DNA. Both parents (I/1 and I/2) harbour two single nucleotide polymorphisms c.1286 C > T and c.1387 G > A localised in exon 8 of the SLC6A5 gene and are therefore heterozygous carriers. Three children, two boys and one girl, were either heterozygous like the parents (II/3) or homozygous (II/2 and II/4) for the GlyT2 mutations. Child II/2 died at the age of 15 months and was homozygous for the c.1286 C > T (P429L at the protein level) mutation but did not carry the c.1387 G > A (D463N at the protein level) polymorphism. Child II/3 harbours both polymorphisms in a heterozygous configuration like the parents and is thus a heterozygous carrier for both DNA polymorphisms. Child II/4 is homozygous for c.1286 C > T but not a carrier of the second c.1387 G > A DNA polymorphism.
[Colour figure can be viewed at wileyonlinelibrary.com] Both GlyT2 variants are listed at the gnomAD database (P429L with only two allele counts out of 251140 alleles (no homozygous), frequency of 0.000008; and D463N with 63540 allele counts out of 282568 alleles (7954 homozygous), frequency of 0.22; https ://gnomad.broad insti tute.org/). The original description of GlyT2 (SLC6A5) as presynaptic component of human hyperekplexia also identified the polymorphism c.1387 G > A (D463N) (Rees et al., 2006). The second child (II/2), a boy, was homozygous for P429L but did not harbour the mutation D463N (Fig. 1a-c). He died at the age of 15 months and suffered from severe hyperekplexia. The third child of the family (II/3) had the same genotype as the parents, representing a carrier of both mutations P429L and D463N (Fig. 1a-c). From the first child (II/1), no material for genotype analysis was available. Due to a lack of symptoms, we expect the same genotype as the parents (marked by the grey square in the pedigree Fig. 1a). The fourth child (II/4), a girl, had the same genotype as the second child (II/2) being homozygous at position c.1286 T (Fig. 1a-c). Again, severe symptoms of hyperekplexia were observed during the first weeks of life. This patient showed generalised stiffness after birth with an exaggerated startle response leading to stiffness with apnoea before treatment with clonazepam was started. At clinical examination, she demonstrated the typical exaggerated head retraction reflex, elicited by tapping her face, especially her chin. During the first year of life, she had repeated episodes of exaggerated startle that ameliorated with increasing doses of clonazepam. At the age of one year, she had mildly delayed developmental milestones and no further apnoeic spells.

| Single DNA polymorphisms are localised on different alleles
To investigate the allele distribution of the two identified polymorphisms, exon 8 was amplified from the genomic DNA of five family members and cloned into the plasmid pRK7 to separate DNA fragments from different alleles (Fig. 2a). From the sequencing data, we expected both mutations on different alleles in the parents (Fig. 2b). Single clones were sequenced, and the frequency of the polymorphisms quantified (Fig. 2c, Table 1). In all cases independent of parent or child, the cloned fragments contained only one of the polymorphisms which is in line with the presence of the two polymorphisms on different alleles. For homozygous patients, all clones contained c.1286 C > T. In contrast, c.1387 G > A was never observed in children suffered from hyperekplexia. The determined allele frequency for the heterozygous child (II/3) was similar to that of the parents (I/1 and I/2) (Fig. 2c,  Table 1).

| GlyT2 P429L shows trafficking through intracellular compartments
Following cloning of the mutation into the cDNA of GlyT2 wild type (wt), the GlyT2 variant P429L was transfected into HEK293 cells and analysed for cellular expression. To perform direct comparison between the GlyT2 wt and the GlyT2 P429L mutant, we analysed transfection capacity of both plasmids compared with co-transfected GFP. The cotransfected GFP as well as the constructs for GlyT2 wt and GlyT2 P429L exhibited similar transfection capacity (GlyT2 wt 25.4 ± 2.4% compared with GFP 26.7 ± 3%; GlyT2 P429L 22 ± 3% compared with GFP 28 ± 2.8%; Fig. 3a,b).
Cells were permeabilized to allow the GlyT2 antibody to reach the intracellular localised epitope. To discriminate between expression in the endoplasmic reticulum (ER) and at the plasma membrane (PM), plasmids encoding fusion proteins of pDsRed and the ER marker calreticulin or the plasma membrane marker GAP-43 were co-transfected. The GlyT2 P429L mutation did not hinder the protein from trafficking (Fig. 4). In co-transfected HEK293 cells, labelling was observed in F I G U R E 2 GlyT2 mutations in the parents are present on two alleles. (a) Allele frequency determined from amplification of exon 8 from five family members (I/1, I/2, II/2, II/3, II/4; except II/1 due to lack of material) followed by cloning and sequencing. (b) Scheme of allele distribution in case of two mutations present on different alleles in the parents (I, paternal and maternal). Mutations on different alleles are labelled with a cross or a circle (black filled). Possibilities of allele division in the second generation (II) with expected frequencies are shown. (c) Quantification of frequency of both mutations (P429L black bars labelled with cross and D463N grey marked with a black circle) from 10 to 12 clones. Note, each clone did always contain one mutation only, either c.1286 C > T (P429L) or c.1387 G > A (D463N) the ER but also at the cellular membrane not obviously distinguishable from the GlyT2 wt protein (Fig. 4a,b). For trafficking analysis through different cellular compartments to the cellular surface, transfected COS7 cells were used. COS7 cells are more suitable for such an analysis due to their large cytoplasm. Similar to the experiments performed in HEK293 cells, GlyT2 wt or GlyT2 P429L was transfected with control plasmids labelling the ER (pDsRed-ER) or the plasma membrane (pDsRed-PM). Moreover, co-labelling with calnexin, an ER marker, ERGIC-53, a marker protein of the ER-Golgi intermediate compartment and GM130, a marker for the cis-Golgi compartment (Fig. 4c,d) resulted in co-localisation of those marker proteins together with either GlyT2 wt or GlyT2 P429L . In conclusion, the mutant as well as the wild-type GlyT2 protein was identified in compartments involved in protein trafficking.
Labelling at the cellular membrane was observed with less intensity for the mutant GlyT2 P429L compared with the GlyT2 wt (lower panels, Fig. 4c,d).

| The GlyT2 mutation results in a significant reduction of whole cell protein expression
The protein biochemical analysis of whole cell lysates from transfected HEK293 cells revealed expression of the GlyT2 wt and GlyT2 P429L . The GlyT2 antibody stained two protein bands at 100 kDa and 75 kDa. The whole cell protein amount obtained from cell lysates was significantly decreased for the mutant GlyT2 P429L compared with the GlyT2 wt protein (51 ± 10%, p = 0.01 compared with GlyT2 wt, Fig. 5a,b). An increase in the loaded protein led to an increase of the signal intensity for both the wt and the mutant protein, but the overall protein amount of GlyT2 P429L was consistently lower compared with the GlyT2 wt protein (Fig. 5c). The single quantification of the mature and immature proteins from the whole cell lysates revealed a significant reduction of GlyT2 P429L with 34 ± 9% (p = 0.033) of the GlyT2 wt for the immature protein (75 kDa) and 61 ± 9% (p = 0.006) for the mature protein (100 kDa) (Fig. 5d). To verify that the mature band (100 kDa) is the glycosylated isoform and the immature protein (75 kDa) is non-glycosylated, membrane preparations obtained from GlyT2 wt and GlyT2 P429L transfected cells were used for digestion with endoglycosidases Endo H and PNGase F (Fig. 5e). PNGase F digestion removes mannose chains directly attached at asparagine. In contrast, Endo H cuts only high-mannose type glycans leaving only one N-acetylglucosamine attached at the asparagine until the Golgi α-glucosidase II cleaves off two mannose molecules. Both, GlyT2 wt and GlyT2 P429L were observed in the cis-Golgi compartment and thus possessed a higher sugar tree capable of being cut by PNGase F (Fig. 5e). Hence, we confirmed that the 100 kDa band refers to the mature GlyT2 and the immature (not fully glycosylated) GlyT2 protein is represented by the 75 kDa as has been reported earlier (Arribas-Gonzalez, de Juan-Sanz, Aragon, & Lopez-Corcuera, 2015; Gimenez et al., 2012).

| GlyT2 P429L is a loss of function mutation
The functional analysis of the GlyT2 P429L mutant was determined from glycine uptake assays. GFP transfection was used as negative control (5.0 ± 0.6 nmol/mg × min, refers to V max ) and represents the endogenous glycine transport of HEK293 cells as has been reported previously (Rees et al., 2006). The transfection efficiencies of GlyT2 wt and GlyT2 P429L together with GFP were similar (Fig. 6a). The glycine transport of GlyT2 wt showed a transport rate of 8.8 ± 0.5 nmol/ mg × min (Fig. 6b, Table 2) which is comparable with previous reports. The GlyT2 P429L mutant lacked glycine transport and was indistinguishable from GFP-transfected cells serving as the negative control (at 250 μM glycine GlyT2 P429L 0.9 ± 0.1 nmol/mg × min; GFP control 1.4 ± 0.2 nmol/ mg × min, non-significant p = 0.39; GlyT2 wt compared with GlyT2 P429L p = 0.0007, GlyT2 wt compared with GFPtransfected cells p = 0.001; Fig. 6b, Table 2). At all glycine concentrations, the GlyT2 wt revealed a significantly increased transport activity compared with the GlyT2 P429L mutant (Fig. 6b,c). To simulate in vitro situation in heterozygous human individuals, GlyT2 wt and GlyT2 P429L were expressed at a 1:1 ratio. The maximal transport activity obtained from the 1:1 wt : mutant expression was not significantly different from the GlyT2 wt expression alone with 6.9 ± 0.4 nmol/mg × min compared with 8.8 ± 0.5 nmol/ mg × min for the wt situation (p = 0.38, Fig. 6b). The remaining transport activity (referring to the maximal transport activity) was 79 ± 5% of GlyT2 wt and explains the nonsymptomatic phenotype of the heterozygous carriers. Using a 10-fold excess of GlyT2 P429L in coexpression with the wild type, the maximal transport activity was 5.5 ± 0.2 nmol/mg x min and thus not significantly different from the mutant GlyT2 P429L (p = 0.41, Fig. 6b, Table 2). The transport activity of the coexpressed GlyT2 wt and GlyT2 P429L mutant in a 1:10 ratio was, however, significantly reduced compared with GlyT2 wt alone (p = 0.0008, Fig. 6b, Table 2). Similar to the overall transport activity, the half-maximal transport activity F I G U R E 5 Whole cell expression differed between GlyT2 wild type (wt) and GlyT2 P429L mutant. (a) Quantification of whole cell GlyT2 wt and mutant protein from lysates normalised to β-actin (both bands were quantified together against β-actin protein amount); wt refers to 100%. (b) Representative gel demonstrating whole cell expression following transfection of GlyT2 wt and GlyT2 P429L into HEK293 cells. GFP plasmid was used for MOCK transfection allowing to control the transfection efficiency in parallel. β-Actin was used as housekeeping protein and stained to verify loading of the same protein amount. The GlyT2 antibody stained two protein bands (100 and 75 kDa). (c) Due to low protein amount determined for the mutant (see the comparison of low and long exposure times), the whole cell lysate was loaded at different protein amounts (5, 10, 20 μg) to the gel. β-Actin served as loading control. (d) Normalized expression of the mutant 100 and 75 kDa protein band to the wt protein (both bands for 100 and 75 kDa were quantified separately, values obtained for wt refer to 100% in each case). Significance values *p ≤ 0.05, **p ≤ 0.01. (e) GlyT2 wt and GlyT2 P429L mutant were digested with endoglycosidases EndoH and PNGase F; non-digested control = ut (untreated). Note, the two bands reflect the mature (100 kDa) and immature (75 kDa) GlyT2 protein, respectively. [Colour figure can be viewed at wileyonlinelibrary.com] was reached at significantly lower glycine concentrations for the GlyT2 wt (239 ± 33 μM) compared with GlyT2 P429L (420 ± 27 μM), and the 1:10 ratio of the GlyT2 wt and GlyT2 P429L (1:10 ratio 393 ± 16 μM) but not different to the 1:1 ratio of wt and mutant (1:1 ratio 272 ± 14 μM, Fig. 6d, Table 2). Hence, 50% of mutated GlyT2 protein does not lead to major changes in the transporter activity; however, 10% of the GlyT2 wt drives the decrease in glycine transport towards lower substrate affinity comparable with the situation in the mutant GlyT2 P429L . Together with the expression analysis, these data reflect that the non-functionality of the mutant GlyT2 P429L is attributed to significantly reduced whole cell GlyT2 P429L protein which lacks glycine transport activity.

GlyT2 mutation
To gain insights into the effect of mutations at the structural level, a homology model of GlyT2 was built and both mutations were mapped onto the structure (Fig. 7a). P429, the site of the P429L mutation resides in α-helix 5 and the second mutation site D463N maps to the loop, which precedes α-helix 6. The in silico mutational analysis suggests that the P429L mutation would not only destroy the kink in the helix produced by the proline but might also affect the mobility of the adjacent α-helix 8. Interestingly, α-helix 8 in the GlyT2 model, corresponding to α-helix-8 in the case of the Drosophila melanogaster DAT structure (Penmatsa et al., 2013) mediates binding of a sodium ion (Na + ). Hence, the altered mobility could potentially play a pivotal role in the gating mechanisms of the transporter. Thus, the P429L mutation can most likely be associated with a sub-optimal functionality of the transporter (Fig. 7b,c). Interestingly, a multiple sequence alignment of human GlyT2, Drosophila melanogaster DAT and prokaryotic LeuT revealed that P429 is conserved in both eukaryotic transporters (Fig. 7f). Inspection of D463 revealed that this residue forms a salt-bridge with K460 located in the short helix at the extracellular region (Fig. 7d,e). The mutation D463N would abolish this salt-bridge while still leaving the ability to form a hydrogen bond.

| DISCUSSION
In this paper, we characterise the pathomechanism of a novel GlyT2 mutation at the molecular, cellular and protein levels generating severe hyperekplexia in affected homozygous individuals. Several reports have shown that mutations in GlyT2 are the second most common source for the neurological disorder startle disease (hyperekplexia) (Carta et al., 2012;Eulenburg et al., 2005;Harvey et al., 2008). Most mutations in GlyT2 identified so far are recessive ranging from missense mutations, frameshift mutations, nonsense or splice site mutations (Carta et al., 2012). In addition to recessive mutants, two dominant negative mutations, Y705C and S512R, were described associated with impaired transporter trafficking and trapping of mutated GlyT2 by calnexin resulting in ER retention (Arribas-Gonzalez et al., 2015;Gimenez et al., 2012). In contrast, recessive missense mutations have not been shown to exhibit disturbed protein transport (Rees et al., 2006). Here, we identified a novel recessive mutation P429L in exon 8 of GlyT2 in a large family with both parents being carriers of two polymorphisms c.1286 C > T and c.1387 G > A. The family had four children, two suffered from startle disease and two unaffected children.
Interestingly, the second polymorphism c.1387 G > A was present also in the unaffected children but never found in the affected children. Thus, we hypothesised that each polymorphism is present on only one allele in the parents as well as the unaffected children. Indeed, we found allelism for both DNA polymorphisms. In case of the two children suffering from startle disease, the polymorphism c.1286 C > T was found homozygous with both alleles affected. The second polymorphism was never detected in the affected individuals when c.1286 C > T was present on both alleles. One child carrying the GlyT2 P429L mutation died at the age of 15 months, the second affected child is being treated with clonazepam, which is usually used as symptomatic treatment. Although the mechanism of clonazepam in this regard is unknown, it is believed that an enhancement of GABAergic inhibition rescues the lack or impairment of the glycinergic inhibition (Andermann, Keene, Andermann, & Quesney, 1980). Due to the severe effect of the mutation GlyT2 P429L in the patient, we setup an experimental series to describe the underlying pathology at the molecular, cellular and functional level. The protein trafficking route was followed through various cellular compartments involved in membrane protein forward transport (Schaefer et al., 2015;Valkova et al., 2011). Usually, ER membrane-bound protein is detectable during polypeptide synthesis (Park & Rapoport, 2012). Furthermore, it was previously shown that GlyT2 glycosylation and interaction with ER chaperones stabilize correct folding and facilitate transporter processing (Arribas-Gonzalez, Alonso-Torres, Aragon, & Lopez-Corcuera, 2013;Martinez-Maza et al., 2001;Rutkevich & Williams, 2011). The subcellular analysis revealed a positive staining of the mutant GlyT2 P429L in the ER colocalized with calreticulin and calnexin, but also with ERGIC53 in the ER-Golgi intermediate compartment and GM130 in the cis-Golgi arguing for ER export of the GlyT2 P429L mutant. Arribas-Gonzalez et al. (2015) reported enhanced accumulation trapped to calnexin for the dominant mutant GlyT2 S521R . Colocalization of GlyT2 P429L with the ER chaperone calnexin as well as detection of the mutant in cellular compartments downstream of the ER and finally surface integration argue for no obvious trafficking defect of the mutant. However, it cannot be | 3917 KITZENMAIER ET Al. excluded that the accumulation with the ER protein calnexin comprises trapping of the GlyT2 P429L protein to some extent which might lead to a migratory defect of the mutated transporter. To further corroborate the GlyT2 P429L mutant protein expression, the whole cell protein of GlyT2 in cell lysates was investigated. GlyT2 was always stained in two protein bands at the respective molecular weights 75 (immature) and 100 (mature) kDa (Arribas-Gonzalez et al., 2015;Gimenez et al., 2012). Cell lysate preparations clearly demonstrated that the general protein expression of the mutant GlyT2 P429L was significantly reduced compared with the wild-type transporter protein. The quantification exhibited around 50% less protein expression. Although the reduced protein expression most probably contributed to the observed disease phenotype in the affected children, a 50% reduction of the mature protein should be sufficient to fulfil the glycine recycling function of the glycine transporter (Carta et al., 2012;Rees et al., 2006). Moreover, GlyT2 P429L was also identified in membranes at the cellular surface. This is in contrast to the dominant negative mutant GlyT2 S512R which showed a complete absence from the cellular surface (Arribas-Gonzalez et al., 2015).

F I G U R E 7
The functional analysis revealed that the mutated protein present at the cellular surface is not capable of transporting the neurotransmitter glycine. The observed transport rate for glycine was indistinguishable from MOCK-transfected cells. The glycine transport efficacy of the GlyT2 wildtype protein was similar to that from other reports in the literature (transport rates between 38-43 nmol/mg × 5 min) (Arribas-Gonzalez et al., 2015;Gimenez et al., 2012;Rees et al., 2006). Since the mutation is recessive, we expected to rescue the transport activity by coexpression with the GlyT2 wild-type protein in a 1:1 ratio (Rees et al., 2006). Indeed, coexpression of GlyT2 wild-type with the mutant in a 1:1 ratio showed no significant difference in the glycine transport activity pattern. This observation matches the symptomless phenotype of the parents whose GlyT2 expression pattern is reflected by a 1:1 ratio of GlyT2 wildtype and mutant protein. Hence, the heterozygous parents and unaffected children sharing the same genotype do not resemble any clinical symptoms of startle disease due to sufficient glycine recycling to enable proper inhibitory neurotransmission. These data clearly demonstrate that the mutation GlyT2 P429L follows a recessive trait which can be rescued by coexpressed wild-type transporter.
An increase in the amount of the mutant protein (90% of mutant GlyT2 P429L protein compared with wild-type protein 10%) shifted the transport activity to lower affinities as observed for GlyT2 P429L suggesting that the loss of function is mostly responsible for the disease pathology. Thus, the presence of two affected alleles resulted in a 51% decrease of the general protein expression, but the remaining GlyT2 P429L transporter lacked protein function.
Finally, we used homology modelling using the Drosophila melanogaster DAT transporter (Penmatsa et al., 2013;Wang et al., 2015) as a template to support the experimental data by identifying possible changes in the structure of the transporter. The location of the GlyT2 mutation P429L in α-helix 5 most probably destroyed the kink in this helix produced by the proline. As a consequence, the mobility of the adjacent α-helix 8 may be impaired. In the structure of the DAT transporter, the location of the P429L mutation is also in α-helix 8 (Penmatsa et al., 2013). In addition to a possible disruption of the α-helical structure of α-helix 5, the amino acid position P429 is close to the sodium-binding site. Hence, the mutation P429L might also affect the gating mechanism of the transporter. In contrast, in silico analysis of the second mutation D463N predicts the loss of a salt-bridge but leaves the ability to form a hydrogen bond. These data are in line with the observed high frequency of the c.1387 G>A polymorphism (D463N at the protein level) in the normal healthy population.
In conclusion, our data expand the current view of the GlyT2 pathology with the identification of a novel missense mutation P429L (homozygous) located in α-helix 5 from a patient with severe startle disease resulting in perturbed glycine recycling and thus a blockade of glycinergic inhibition.