Isoform Heterogeneity of the Human Gephyrin Gene (GPHN), Binding Domains to the Glycine Receptor, and Mutation Analysis in Hyperekplexia*

From the Departments of ‡Molecular Medicine and ¶¶Anatomy, Faculty of Medical and Health Sciences, University of Auckland, Private bag 92019, Auckland, New Zealand, the ¶Department of Pharmacology, School of Pharmacy, 29-39 Brunswick Square, London WC1N 1AX, United Kingdom, Pathway Discovery, Genomics and Proteomic Sciences, GlaxoSmithKline Medicines Research Centre, Gunnels Wood Road, Stevenage SG1 2NY, United Kingdom, **University of Birmingham Medical School, Birmingham B15 2TT, United Kingdom, the §§Department of Psychological Medicine, University of Wales College of Medicine, Heath Park, Cardiff CF14 4XN, United Kingdom, the Department of Paediatric Neurology, University Hospital of Wales, Heath Park, Cardiff CF14 4XN, United Kingdom


Gephyrin (GPHN) is an organizational protein that clusters and localizes the inhibitory glycine (GlyR) and
GABA A receptors to the microtubular matrix of the neuronal postsynaptic membrane. Mice deficient in gephyrin develop a hereditary molybdenum cofactor deficiency and a neurological phenotype that mimics startle disease (hyperekplexia). This neuromotor disorder is associated with mutations in the GlyR ␣ 1 and ␤ subunit genes (GLRA1 and GLRB). Further genetic heterogeneity is suspected, and we hypothesized that patients lacking mutations in GLRA1 and GLRB might have mutations in the gephyrin gene (GPHN). In addition, we adopted a yeast two-hybrid screen, using the GlyR ␤ subunit intracellular loop as bait, in an attempt to identify further GlyR-interacting proteins implicated in hyperekplexia. Gephyrin cDNAs were isolated, and subsequent RT-PCR analysis from human tissues demonstrated the presence of five alternatively spliced GPHN exons concentrated in the central linker region of the gene. This region generated 11 distinct GPHN transcript isoforms, with 10 being specific to neuronal tissue. Mutation analysis of GPHN exons in hyperekplexia patients revealed a missense mutation (A28T) in one patient causing an amino acid substitution (N10Y). Functional testing demonstrated that GPHN N10Y does not disrupt GlyR-gephyrin interactions or collybistininduced cell-surface clustering. We provide evidence that GlyR-gephyrin binding is dependent on the presence of an intact C-terminal MoeA homology domain. Therefore, the N10Y mutation and alternative splicing of GPHN transcripts do not affect interactions with GlyRs but may affect other interactions with the cytoskeleton or gephyrin accessory proteins.
Neuronal postsynaptic membranes contain a wide variety of ion channels and receptor proteins that contribute to the balance of excitatory and inhibitory neurotransmission in the human central nervous system. Mature postnatal neuroinhibition is mediated by postsynaptic glycinergic and GABA A 1 receptor systems that are structurally related to the nicotinic receptors with a typical pentameric structure constructed from heterogeneous subunit constituents containing four transmembrane (TM) domains and a large intracellular loop between TM3 and TM4 (1,2). These subunits combine to form fast response, ligand-gated chloride channels that modify excitatory neurotransmission in the human brainstem and spinal cord (3)(4)(5). Inhibitory glycine receptors (GlyRs) are composed of three ligand-binding ␣ 1 subunits (GlyR ␣ 1 ) and two structural ␤ subunits (GlyR ␤) that are anchored and clustered on the postsynaptic membrane of inhibitory neurons by the interaction of a motif within the TM3-TM4 loop of the GlyR␤ subunit with anchoring protein, gephyrin (6 -8). In contrast to GlyRs, there is a lack of evidence for a direct physical interaction between gephyrin and GABA A receptor subunit(s) (7,9), although immunohistochemical studies have detected co-localization of gephyrin with major subtypes of GABA A receptors at GABAergic postsynaptic sites in murine and rat brain (10 -14).
Gephyrin (GPHN) belongs to a group of cytoskeletal elements that are critical for receptor-associated synaptic localization and organization. First identified as a 93-kDa protein by co-purification with GlyRs from rat spinal cord, GPHN demonstrated high affinity binding to tubulin (15)(16)(17). It is suggested that gephyrin forms a postsynaptic organizational lattice structure that dynamically immobilizes and clusters inhibitory receptors onto the cytoskeletal infrastructure (14,18). In support of this, GPHN-dependent postsynaptic clustering of GlyRs was demonstrated by cellular antisense experi-ments and from studies of a gphn knock-out mouse model (12,19,20). Heterogeneous roles for gephyrin have been implicated by protein-protein interactions with a range of determinants such as RAFT1, tubulin, profilin, collybistin, and the dynein light chains 1 and 2 (17,(21)(22)(23)(24).
The preliminary structure of the gephyrin gene (GPHN) has been described in mice and humans (25)(26)(27), although there is a degree of ambiguity in the number of exons due to the presence of multiple transcript isoforms, and it is uncertain that all exons have been located and identified. Indeed, GPHN expression is not restricted to rat brain and spinal cord but is also found in liver, kidney, lung, and retina (28 -30). Gephyrin knock-out mice and reconstitutive cell culture assays have demonstrated that gephyrin expression in nonneuronal tissue is a requirement for the biosynthesis of molybdenum cofactor (20,31). Striking homologies are observed between the primary structure of GPHN and proteins involved in bacterial, plant, and invertebrate molybdenum cofactor biosynthesis. Gephyrin, therefore, has a synaptic function as a postsynaptic anchoring and clustering protein in neurons while facilitating a highly conserved metabolic purpose in nonneuronal tissues (32). The genetic and structural basis of the functional dichotomy is not established; however, in rats, a putative explanation may lie in the generation of distinct transcript isoforms of gephyrin constructed from alternative splicing of eight exonic "cassettes" within four regions of GPHN (25,28,33). Apart from conferring neuronal and nonneuronal characteristics to the mature polypeptide products, the purpose of the transcript heterogeneity pattern is unclear, and the number of isoform combinations in human tissue remains unestablished.
In addition to major disruption of molybdenum cofactor biosynthesis, gephyrin-deficient mice display a neuromotor phenotype, which resembles human hereditary hyperekplexia. This neurological condition (Mendelian Inheritance of Man: 149400 and 138491) can be caused by dominant and recessive mutations in the GLRA1 and GLRB receptor genes and is characterized by an abnormal, persistent startle response to unexpected stimuli, neonatal hypertonia, and a chronic accumulation of injuries caused by unprotected, startle-induced falls (34 -40). Furthermore, autoimmunity to GPHN was detected in a patient with Stiff-Man syndrome, a disorder that has a degree of phenotypic overlap with hyperekplexia (41). Collectively, this indicates that GPHN is a candidate gene for neurological and metabolic disorders, and support for the latter has recently emerged with a description of a GPHN reading frame mutation in a family with hereditary molybdenum cofactor deficiency (42).
In response to the candidacy of gephyrin for hyperekplexia, we pursued the definitive genomic structure of gephyrin by a combination of in silico BAC contig construction and multitissue RT-PCR methods. Using a yeast two-hybrid (YTH) screen, with the GlyR ␤ subunit intracellular loop as bait, we attempted to identify further GlyR-interacting proteins that might be implicated in hyperekplexia. The results of this screen, together with the extensive search for all human GPHN transcripts in both neuronal and nonneuronal tissues, suggest the presence of one new GPHN exon and at least 11 neurological transcript isoform combinations that redefine the genomic organization of GPHN. Having established the number of exons in the GPHN gene, we completed a systematic exon by exon mutation analysis of GPHN in a cohort of hyperekplexia patients devoid of GLRA1 and GLRB mutations. One novel mutation (N10Y) and several SNPs are reported; however, the functional effect of the mutation remains elusive despite assays for receptor targeting and clustering. We also present evidence that GlyR-gephyrin binding is dependent on the presence of an intact C-terminal MoeA homology domain and that the heterogeneous GPHN linker region is not the physical determinant for gephyrin-GlyR binding. (Table I) was used in a first-strand cDNA synthesis reaction (SuperScript II; Invitrogen) from total RNA derived from adult brain, fetal brain, cerebellum, and human spinal cord (Clontech). In addition to neurological tissue, cDNA was synthesized from a selection of nonneurological tissues including heart, liver, lung, kidney, trachea, fetal liver, pancreas, and placenta (Clontech). In the first instance, exonic primers were designed in constitutional regions surrounding the candidate regions of alternative splicing, namely regions C1/2, C3, C4/5, and C6/7 (Table I). To validate the initial RT-PCR methodology, downstream nested primers within the GPHN spliced exons were used in conjunction with primers from upstream invariant portions of the GPHN message (Table I). In the presence of multiple PCR products, DNA fragments were cloned into pGEM-Easy vectors (Stratagene), and the transformants were PCR-screened and assessed for size differences by molecular screening agarose (Roche Applied Sciences) and 10% nondenaturing polyacrylamide gels (Sigma). Minipreparations of clones were digested with EcoRI, and inserts were sequenced using ABI 3100 technology.

RT-PCR Analysis of GPNH Alternative Splicing-A GPHN genespecific primer
YTH Screening-To identify GlyR ␤ subunit interactors in human brain, we cloned the large intracellular loop of the human GlyR ␤ subunit into the plasmid vector pYTH9 (43,44). In this manner, the large TM3-TM4 intracellular loop (known to harbor a gephyrin binding motif) (8) was fused to the GAL4 DNA binding domain. pYTH9 is advantageous for YTH screening, since the bait plasmid can be stably integrated into the yeast genome at the trp-1 locus. An adult human brain cDNA library in pACT2 (HL4004AH; Clontech) was be screened by transformation of library DNA into the pYTH9-GlyR ␤ yeast strain. Transformed yeast were plated on selective dropout media lacking leucine, tryptophan, and histidine, supplemented with 10 mM 3-amino-1,2,4-triazole, a competitive inhibitor of the HIS3 gene product. Transformation plates were incubated at 30°C for 12 days to allow histidine prototropic colonies to emerge. Surviving colonies were restreaked onto selective agar, lifted onto Whatman 54 paper, and screened for proteinprotein interactions using a standard freeze-thaw fracture assay for the lacZ reporter gene (44). Library plasmid DNAs were recovered from ␤-galactosidase-positive colonies using the Yeastmaker plasmid isolation kit (Clontech), transformed into Escherichia coli XL1-blue (Stratagene), miniprepped, and sequenced. True interactions were checked by retransformation of library plasmids into yeast containing pYTH16 (an episomal version of pYTH9) and pYTH16-GlyR ␤. Library plasmids that were capable of autoactivation (i.e. activated lacZ with pYTH16 alone) were excluded from further analysis.
In Silico Determination of Human Gephyrin Gene Structure-Rat cDNA (28) and human expressed tag sites similar to rat gephyrin were submitted for BLAST searches (available on the World Wide Web at www.ncbi.nlm.nih.gov/BLAST/). Portions of rat cDNA fragments displayed 90 -95% homology with dispersed regions of several chromosome 14 human genome BACs submitted to GenBank TM . Exon/intron organization was established from the genomic BAC contig, and a putative promoter region was identified through sequence homologies. Details of the human GPHN gene have subsequently been described by David-Watine (27), and the BAC clones are accessible from GenBank TM .
Hyperekplexia Patients-The majority of patients included in the mutation analysis of the GPHN gene (n ϭ 31) are described elsewhere (37). In addition, a further seven unrelated hyperekplexia patients all exhibited to the diagnostic criteria of inclusion, which involves a history of neonatal hypertonia, a nose tap response, and an exaggerated startle response, leading to injurious fall down consequences with preservation of consciousness.
PCR Analysis of GPHN in Hyperekplexia-The GPHN promoter region, exons and flanking intronic sequences, and 5Ј-and 3Ј-untranslated regions were amplified from patient DNA. Primer sets were designed by Primer 3.0 criteria via the online facility (available on the World Wide Web at www-genome.wi.mit.edu/cgi-bin/primer/primer3.0). SSCP-Samples for screening were prepared by denaturing 5 l of GPHN PCR products with 7 l of formamide dye at 94°C for 5 min. Samples were cooled and applied to 10% nondenaturing gels (49:1; Sigma) and run at 75 V for 12-16 h at room temperature and 4°C. All SSCP gels were silver-stained as previously described (47).
dHPLC-Analysis was carried out using the Transgenomics dHPLC 2100 WAVE R DNA Fragment Analysis System and DNASep R column (Transgenomic, Santa Clara, CA). The dHPLCMelt program was used to predict the optimal melting conditions for PCR fragments, which were analyzed on a WAVE DNA fragment analyzer at temperatures recommended by predicted helicity profiles across the DNA fragment. Variant dHPLC profiles suggestive of sequence heterogeneity were sequenced. The WAVE was run under partially denaturing conditions for mutation detection and SNP discovery. Wavemaker 4.0 software was used to control the rate of flow of buffers and column running. A volume of between 4 and 8 l of patient and control PCRs were injected into the column depending on the DNA concentration of the PCR.
DNA Sequencing-GPHN SNPs were detected by radionucleotide [␣-33 P] cycle sequencing. Sequencing was initiated using a denaturation stage at 94°C for 5 min followed by 45 cycles of primer annealing temperatures at 55°C for 30 s, primer extension at 72°C for 1 min, and denaturation at 94°C for 30 s. Denatured samples were applied to a 6% denaturing polyacrylamide gels (Sigma) and resolved at 85 watts for 1-2 h at room temperature. The sequence content of RT-PCR products and clones were determined by ABI 3100 equipment using BigDYE chemistry (PerkinElmer Life Sciences). Each amplimer was sequenced with sense and antisense primers using an ABI PRISM TM BigDye Terminator cycle sequencing kit and AmpliTaq DNA polymerase (PerkinElmer Life Sciences).
Population Studies-The population frequency of identified sequence variations (potentially pathogenic mutations and SNPs) were established in a variable number of unrelated controls of Caucasian descent and derived from blood donor clinics in New Zealand (Table II). In instances where a restriction site change was predicted by the polymorphism, the relevant restriction fragment length polymorphism assay was designed and resolved by agarose gel electrophoresis or polyacrylamide gel electrophoresis. For those SNPs devoid of restriction site changes, the SNP frequency in the normal population was determined by comparison of mutation profiles versus normal profiles using mutation screening techniques SSCP, DDF, or dHPLC (see Table II).

RESULTS
GlyR ␤ Subunit YTH Screen-In an attempt to identify further GlyR-interacting proteins that may be implicated in hyperekplexia, we carried out a YTH screen of an adult human brain cDNA library, using the GlyR ␤ subunit TM3-TM4 intracellular loop as bait. The cDNA library had an original complexity of 3.5 ϫ 10 6 independent cDNA clones, and 8 ϫ 10 6 cDNAs were screened. Twelve positive clones were recovered, which specifically interacted with the GlyR ␤ subunit bait. These represented eight independent cDNAs encoding variants of the human homologue of GPHN (Fig. 1). No other interacting proteins were found. Some of our gephyrin cDNAs (e.g. Geph1) were full-length, which allowed us to predict the entire coding sequence of the human gephyrin polypeptide. Structurally, the N-and C-terminal domains of gephyrin show high similarity to E. coli proteins (MogA and MoeA), a Drosophila protein (CIN-NAMON), and an Arabidopsis thaliana protein (CNX1), all of which are involved in the synthesis of a molybdenum-containing co-factor essential for molybdoenzyme activity. Analysis of the phylogenetic conservation within the human GPHN open reading frame reveals a 94% homology to rodent gphn sequence and a 91% homology to chicken sequence, assuming the exclusion of all alternatively transcribed exons. In comparison with rodents, this is translated into a 99.7% amino acid conservation (Fig. 2) within the constitutional 736-amino acid reading frame. Degeneracy is observed at two sites, I240L and A242R, amino acid changes that lie immediately 5Ј adjacent to the C3 linker insertion site (30). The amino acid conservation between chicken and human GPHN is reduced to 98.7%, with all but one of the nine degenerate changes occurring at the MoeA domain, and a reversal of the rodent I240L and A242R changes is observed (Fig. 2).
Interestingly, several of the GlyR ␤-interacting GPHN clones (Geph5-8) were missing all of the N-terminal MogA homology domain (amino acids 16 -169) as well as most of the "linker" segment of gephyrin (amino acids 170 -322; Fig. 1). This suggests that an intact C-terminal MoeA domain is required for FIG. 1. Gephyrin clones captured by yeast two-hybrid screen. Shown is a schematic diagram displaying the domain structure of the human gephyrin P1 isoform (GephP1) indicating the positions of the MogA and MoeA homology domains (amino acids 16 -169 and 323-736) and the intervening "linker" segment (amino acids 170 -322). Below this are the structures of the human gephyrin cDNAs isolated in our GlyR ␤ subunit YTH screen (Geph1-Geph8). Note that some cDNAs contained either C3 (Geph2) or single (Geph3) or multiple (Geph1) C4 cassettes. In addition, all cDNAs encode an intact MoeA homology domain, suggesting that this region is essential for mediating robust interactions with the GlyR ␤ subunit. strong interactions with the GlyR␤ subunit and that the binding site on gephyrin for GlyRs does not reside in the MogA homology domain or the linker region.
Transcript Heterogeneity of Human GPHN-Several variants of gephyrin were found in our GlyR ␤ subunit YTH screen. That gephyrin exists in multiple isoforms has been established for some time (25,28,33); the most common variants have insertions in the "linker" region, where exon C3 inserts the peptide NHPFYTSPAVFMANHGQPIPGLISYSHHATGSA-DKR between amino acids 243 and 244, and the C4 exon inserts the sequence ARLPSCSSTYSVSE between amino acids 288 and 289. More recently, Ramming et al. (25) described seven alternatively expressed murine exons including three consecutive C4 exons and two further 5Ј exons designated C6 and C7. In our Y2H screen and in subsequent RT-PCR experiments (Fig. 3), we have found that human GPHN has at least four exons within the C4 cassette cluster; C4A, QIRRPDESKGVASRVGSLK (identical to original C4b cassette) (50); C4B, LHRKLEELRDHLEGNVKGYSLRVNVRFHL (novel exon); C4C, ARLPSCSSTYSVSE (identical to the original C4 cassette) (28); and C4D, LHSRLEGLKDELWRSRG-YDLR (identical to the original C5 cassette) (25). The C4A-C-4B-C4C-C4D cluster is physically contiguous on human chromosome 14 BACS and definitively alters the genomic organization of GPHN (see Fig. 4).
The RT-PCR assays for human GPHN were designed to capture regions of splicing activity implicated from rodent gphn studies (Table I). Twelve different human RNA preparations from neurological (n ϭ 4), and nonneurological (n ϭ 8) tissues were tested (see Fig. 3A). The presence of rat gphn exons C1 (region 1) or exon C5* described by David-Watine (27) was not detected in any of our transcripts from human tissues. Exon C2 (region 1) was ubiquitously expressed in all transcripts with no suggestion of differential splicing. Furthermore, murine gphn exons C6 and C7 were not detected; nor was the exclusion of exon 18 in human GPHN transcripts (25, 51). The transcript pattern surrounding the C3 and C4 cluster revealed a clear difference between neurological tissues and nonneurological tissues. A solitary transcript was found in all nonneurological tissues, which included the C3 exon and excluded all exons from the C4 cluster (Fig. 3A). In contrast, neurological tissue generated a ladder effect, indicating the presence a range of fragments. The fragments were cloned into pGEM-easy vectors, and a total of 547 clones were PCR-screened for size differences, from which 86 clones were selected and sequenced. All of the clones sequenced contained perfectly preserved splicing insertion points and followed the linear 5Ј to 3Ј order of exons on the BAC contig. In total, sequencing confirmed the presence of 11 distinct neurological GPHN isoforms (Fig. 3B). The majority of neurological isoforms excluded exon C3 and contained a combination of the C4 exon cluster; however, two rare clones were detected that demonstrated the co-inclusion of exon C3 and one of the C4 cluster exons (Fig. 3B). One isoform corresponded to the nonneurological pattern by exclusively retaining the C3 exon, and this presumably reflects the metabolic transcript of gephyrin required in all tissues including the human brain.
To validate our capture of all transcript isoform combinations, further RT-PCRs were generated from an upstream invariant forward primer and nested reverse primers within each spliced exon of the C4 cluster (Table I). This approach captured all previously identified isoforms and did not reveal any additional isoform combinations. Finally, additional testing of the remaining GPHN invariant regions (Table I) found no evidence to suggest any exonic splicing outside the C3 and C4 exon cluster.
Mutation Analysis and SNP Distribution in GPHN-Three highly informative mutation detection techniques (SSCP, dideoxy fingerprinting, and automated dHPLC) were adopted to detect GPHN mutations and polymorphisms in a cohort of 32 hyperekplexia patients. Several SNPs were detected in the promoter, intronic, and exonic regions in both patient and control cohorts (Table II). In addition, a 4-bp deletion was detected within the GPHN candidate promoter region (del tgga at position Ϫ976 3 Ϫ972), which was also detected in normal controls and was not functionally assessed. Significantly, a heterozygous missense mutation (A28T) was detected in a hyperekplexia patient that causes an asparagine to tyrosine amino acid substitution (N10Y) in exon 1 of the GPHN gene (Fig. 5A). There were no restriction site changes for GPHN A28T, so consequently we analyzed 94 controls by DDF and confirmed the exclusion of patient-specific patterns from the control group (Fig. 5B). The N10Y substitution is positioned 5 amino acids upstream from a putative myristoylation site, which is a conserved marker for protein interactions with a membrane-bound receptor complex or the lipid bilayer of the plasma membrane. It was proposed that the N10Y mutation disturbed this interaction, thereby compromising the GPHN lattice at an anchorage point, leading to a disruption in GlyR Only the C3 and C4 assays indicated the presence of transcript heterogeneity. B, size-ordered GPHN-pGEM clones derived from a C3/C4 cluster combination assay and digested with insert-releasing EcoRI restriction enzyme and resolved on a 10% polyacrylamide gel. In total, 11 neuronal GPHN isoforms were discovered, and each isoform corresponded to the expected sizes from the possible exonic combinations within the PCR assay. Each of the spliced exons are represented with distinct patterns, and amino acid translations are provided. Exons in the C4 cluster have been assigned a new nomenclature, and a conversion key relating to the old nomenclature is provided. Clone 4 was the sole transcript observed in nonneurological tissues (n ϭ 8).
clustering and reduced attenuation of the startle response.
Functional Analysis of GPHN N10Y-To assess whether the mutation N10Y could affect interactions between GPHN with the GlyRs or with the GDP/GTP exchange factor collybistin, we carried out functional expression experiments in transfected human embryonic kidney cells (HEK293). We used expression constructs encoding (i) the GlyR ␤ subunit intracellular loop tagged at the C terminus with red fluorescent protein (DSRed-GlyR ␤), (ii) collybistin II (CB2SH3-) tagged at the N terminus with the 9E10 (Myc) epitope, and (iii) gephyrin P1 wild-type and N10Y mutants tagged at the N terminus with EGFP. As expected, DSRed-GlyR ␤ and Myc-CB2SH3 fusion proteins showed cytoplasmic expression (Fig. 6). In contrast, EGFPgephyrin (not shown) and EGFP-gephyrin N10Y formed small cytoplasmic aggregates that readily trap DSRed-GlyR ␤ (Fig.   6). Since EGFP-gephyrin N10Y is also translocated to the cell surface by collybistin II, it is unlikely that the N10Y mutation affects the structural lattice formed by gephyrin or interrupts gephyrin-GlyR ␤ or gephyrin-collybistin interactions. These findings are reinforced by our yeast two-hybrid data, which strongly suggest that the GlyR ␤ binding site on gephyrin is located within the C-terminal MoeA homology domain. DISCUSSION The pleiotropic nature of GPHN demands a molecular explanation, not only to account for the functional differences between neurological and nonneurological tissues but also to understand the basis for preferential recruitment and targeting of neurotransmitter receptors to the postsynaptic membrane. Derived from an extensive RT-PCR screen of the GPHN   FIG. 4. A revised genomic organization of human GPHN. The previous version of the human GPHN gene described the existence of 27 exons distributed over 760 kb of genomic DNA on chromosome 14q32 (Fig. 2A). This remains true as a summary; however, our transcript data have led to a change in the order of exons, the exclusion of human exon C5*, and exclusion of rat C1, C6, and C7 from human GPHN transcripts. Based on these data, we suggest a new genomic representation of GPHN and a numbering system for all constitutional and spliced exons (Fig. 3B). We further recommend co-retaining the "C" nomenclatures for the C3 region and the (C4b-C4 new -C4-C5) cluster, with revision of the latter to the more logical labeling of C4A-C4B-C4C-C4D, reflecting the physical genomic organization of these exons within human GPHN.

TABLE I RT-PCR primers for the spliced regions of GPHN
These primers were used in the generation and capture of human GPHN transcript isoforms. A GPHN-specific primer located at the 3Ј-end of the gene was used to generate cDNA fragments from mRNA templates derived from neurological (n ϭ 4) and nonneurological (n ϭ 8) tissues (Clontech). Primer sets for each candidate region of alternative splicing were designed in regions of the gene that were invariant from tissue to tissue. In addition, primer sets were designed to analyze the complete reading frame in order to exclude any further regions of exon splicing (Invariant 1 and 2). The exonic origin of each primer is shown in superscripts. Four nested primers located within each spliced exon were used in validation experiments to ensure capture of all GPHN transcript isoforms. UTR, untranslated region.  gene and YTH clone capture, we provide evidence for extensive alternative intense splicing activity in neuronal tissue, which was in contrast to the single transcript observed in all eight nonneuronal tissues. The single nonneurological transcript includes exon C3, an exon that is absent from the majority of neuronal-specific transcripts, although human brain contains a mosaic expression of metabolic nonneuronal and heterogeneous neuron-specific isoforms. We suggest that the nonneuronal gephyrin transcript is ubiquitous in human tissues and serves a metabolic housekeeping purpose as a cofactor in molybdenum enzyme synthesis. The substantial combinatory nature of the C4 exon cluster is created from four alternatively spliced exons, which generate 10 neuronal specific isoforms. The majority of these transcripts do not retain exon C3; however, two rarely detected isoforms have retained exon C3 and one of the C4 cluster exons. Three of the four exons observed in the C4 region have been reported in rodent tissues (25,28). One novel exon was detected in human transcripts and is physically located, in the correct linear order, within the chromosome 14 BAC contig. Collectively, with the specific exon to exon construction pattern within all sequenced isoforms and the BAC contig validation, we are confident that the 5Ј 3 3Ј order of the C3 and C4 spliced exons is correct. There remains a possibility that not all GPHN isoforms have been captured because of the sheer number of combinatory possibilities; however, any additional isoforms may be extremely rare and transcribed in specific areas of the human brain. In phylogenetic terms, our lack of detection for rodent exons C1, C6, and C7 in human tissue is also noted by David-Watine (27). It remains to be seen whether these exons exist, whether they are so weakly and specifically expressed that we did not detect them or whether they are an experimental artifact. For a gene that is so highly conserved, it is surprising that entire exons have been disregarded during recent evolution, but further work in rodents and human brain sections may be required to resolve this issue. It is significant, however, that rodent exons C6 and C7 are not detected, in a conserved or degenerate manner, within a small region of the chromosome 14 BAC contig containing the expected insertion points between exons 15 and 19 of GPHN. All remaining 27 human GPHN exons (including C3 and C4 cluster) are easily detectable on the BAC contig by BLAST comparison with homologous rodent exons. In summary, the YTH and PCR results demonstrate that (i) the complexity of alternative splicing of gephyrin has been greatly underestimated; (ii) a revised genomic structure for human GPHN and a simplified nomenclature for the exons within the C4 cluster is required; (iii) the GlyR ␤ subunit is capable of interacting with multiple gephyrin isoforms; and (iv) molecular mechanisms exist whereby gephyrin could regulate complex associations with different cytoskeletal or cytoplasmic proteins at glycinergic versus GABAergic synapses.
The speculative functional consequence of the heterogeneous GPHN isoforms is attributed to interactions with a complex repertoire of protein interactions. The present list includes FIG. 5. Mutation analysis of GPHN in hyperekplexia. A, sequence analysis of GPHN exon 1 in a hyperekplexia patient using ␣-33 P cycle sequencing. A synonymous heterozygous change (A28T) was observed in exon 1 of GPHN, resulting in an N10Y amino acid substitution, positioned 6 residues upstream of a putative myristoylation site (28). B, a nondenaturing, 10% MDE gel illustrating a population study to compare patient GPHN A28T and population control DDF patterns. A patient-specific pattern (indicated by arrows in columns labeled P) was not detected in unrelated Caucasian control samples (n ϭ 94; samples 1-28 illustrated).
FIG . 6. Functional assessment of GPHN-N10Y. The expression assays demonstrate that the GPHN N10Y mutation is unlikely to interfere with gephyrin-GlyR ␤ subunit or gephyrin-collybistin interactions. When expressed alone in HEK293 cells, the GlyR ␤ subunit intracellular loop tagged with red fluorescent protein (A; DSRed-GlyR ␤) and Myc-tagged collybistin II (C; Myc-CB2SH3-) are found throughout the cell. In contrast, note that EGFP-gephyrin N10Y forms cytoplasmic aggregates (B), which are identical to those formed by EGFP-gephyrin. EGFP-gephyrin N10Y readily traps DSRed-GlyR ␤ in these cytoplasmic aggregates (D-F) and is translocated to the cell surface upon co-expression with collybistin II (G-I). The results reveal no significant differences compared with control experiments performed using nonmutated gephyrin (i.e. EGFP-gephyrin; data not shown). evidence for protein interactions with the ␤ subunit of the glycine receptor, tubulin, GABARAP, profilin, collybistin, and the dynein light chains. Considering that this is probably not the complete cast of interacting proteins, it becomes apparent that the splicing activity of C3 and C4 regions of the GPHN gene may form the basis for differential properties of GPHN isoforms. The "linker" segment also contains several potential protein-protein interaction domains including proline-rich and acidic segments as well as a short motif (amino acids 289 -299) that has strong similarity to the tubulin-binding domain signature of tau and MAP2/MAP4 proteins (25). This may be of significance, since the tubulin-binding properties of gephyrin closely resemble those of mitogen-activated proteins (52). Given the proximity of the site of insertion for the C4 cassette to the putative tubulin-binding domain, it is possible that C4 inserts might influence interactions with cytoskeletal proteins.
Interestingly, cassettes C4B and C4D are homologous to each other and to the ␣-helical region 1B of human keratin 8 (53). This motif is known to mediate keratin dimerization by the formation of coiled-coil structures. It is possible that inclusion of cassettes C4B and or C4D creates a binding site for a new gephyrin interactor. Ramming et al. (25) concluded that gephyrin transcripts containing the C3 and C4 exons were of low abundance but highly expressed in some subpopulations of neurons in the adult mouse brain. For example, cassette C4D was found in the granular cell layer of the cerebellum and olfactory bulb. However, this study did not take other developmental stages into account (e.g. cassette C3 is highly expressed throughout the central nervous system early in development) and did not focus on spinal cord or brainstem. Generating spliced exon-specific antibodies and immunohistochemical analysis in rodent and human brain will greatly enhance our understanding of spatial and temporal distribution of gephyrin transcripts.
Despite the established association of hyperekplexia mutations with subunits from glycinergic ion channels in both humans and animal models, it is apparent that mutations in gephyrin are not a major cause of the disorder. Considering the high degree of phylogenetic conservation and that GPHN is an essential metabolic agent, it is possible that mutations in many regions of GPHN are embryonic lethal. Our promoter and exon by the exon mutation detection approach revealed a novel mutation causing a N10Y substitution at the extreme N terminus of GPHN. This portion of gephyrin is not part of the MogA consensus sequence, which begins at amino acid 15, and this positioning may be the basis of functional tolerance in the mutation carrier. Recent crystal structure determination of the GPHN N-terminal domain indicates a trimeric structure that presumably is the structural conformation required to form a subsynaptic lattice (54,55). The N10Y mutation was found in a young male infant who presented with neonatal hypertonia and excessive startle response in the first year of life. At 4 years of age, there has been a transient recovery of the phenotype with no features of molybdenum metabolism disruption. GPHN N10Y may have a limited detrimental window during developmental glycinergic switching from GlyR ␣ 2 homopentamers to GlyR ␣ 1␤ heteropentamers. From a glycinergic perspective, GPHN N10Y does not disrupt GlyR-gephyrin interactions in human embryonic kidney cells and does not interrupt collybistin-induced cell surface clustering. Since in vitro clustering is not affected by GPHN N10Y , the functional effect of this mutation remains elusive, and an alternative mechanism is responsible for the hyperekplexia phenotype.
The fact remains, however, that ϳ60% of our sporadic hyperekplexia cohort is devoid of GLRA1, GLRB, and GPHN mutations, and this clearly implies further genetic heterogene-ity. Casting a wider net for the mutation analysis of glycinergic transporters and GABAergic candidates may yet yield further genetic heterogeneity in hyperekplexia. Indeed, recent descriptions of GABA A mutations in epilepsy demonstrates the subtle functional overlap between the two neuroinhibitory systems with the resulting phenotypes a reflection of the differential localizations and function of glycinergic and GABAergic receptor subunit expression.
In conclusion, we have demonstrated the existence of a large number of neuron-specific gephyrin isoforms that may form the basis for specialized protein interactions. By identifying all exon constituents of the gene, we were able to exclude the gephyrin coding sequence as a major determinant for hyperekplexia. Nevertheless, a solitary mutation may yet have a functional explanation in a transient form of the disorder despite the exclusion of our initial functional hypothesis. We describe a number of synonymous and flanking intronic SNPs of GPHN that may be informative for mapping purposes and other disorders linked to chromosome 14q24. Finally, as a consequence of a YTH search for glycinergic interacting proteins, we can confirm that the physical location of the GPHN-GlyR ␤ interacting domain in the GPHN molecule does not lie in the MogA region, linker segment, or C3 and C4 splicing cluster. This domain lies within the C-terminal portion of the conserved MoeA region, and the definitive identification of an interacting GPHN motif represents the next challenge toward characterizing the physical nature of GPHN-GlyR coupling.