The Human Glycine Receptor Subunit a3 GLRA3 GENE STRUCTURE, CHROMOSOMAL LOCALIZATION, AND FUNCTIONAL CHARACTERIZATION OF ALTERNATIVE TRANSCRIPTS*

The neuronal glycine receptor is a ligand-gated chloride channel composed of ligand binding a and structural b polypeptides. Homology screening of a human fetal brain cDNA library resulted in the identification of two alternative splice variants of the glycine receptor a3 subunit. The amino acid sequence predicted for the a3L variant was largely identical to the corresponding rat subunit. In contrast, the novel splice variant a3K lacked the coding sequence for 15 amino acids located within the cytoplasmic loop connecting transmembrane spanning region 3 (TM3) and TM4. Using P1 artificial chromosome (PAC) clones, the structure of the GLRA3 gene was elucidated and its locus assigned to human chromosomal bands 4q33-q34 by fluorescence in situ hybridization. Two transcripts of 2.4 and 9 kilobases, corresponding to a3L and a3K, respectively, were identified and found to be widely distributed throughout the human central nervous system. Structural analysis of the GLRA3 gene revealed that the a3K transcript resulted from a complex splice event where excision of the novel exon 8A comprising the alternative sequence of 45 base pairs coincides with the persistence of a large intronic sequence in the 3*-untranslated region. Functional expression in HEK 293 cells of a3L and a3K subunits resulted in the formation of glycine-gated chloride channels that differed significantly in desensitization behavior, thus defining the cytoplasmic loop as an important determinant of channel inactivation kinetics.

Glycine serves as a major inhibitory neurotransmitter throughout the mammalian central nervous system (1). The strychnine-sensitive glycine receptor (GlyR) 1 is a pentameric assembly of ligand binding ␣ and structural ␤ subunits displaying significant sequence homology to nicotinic acetylcholine receptor (2), ␥-aminobutyric acid receptor type A (GABA A R), and serotonin receptor type 3 (5-HT 3 R) subunits (1). As members of a superfamily of ligand-gated ion channels, these polypeptides share topological features including a large Nterminal extracellular domain followed by four transmembrane spanning regions (TM1-TM4). While the N-terminal domain carries structural determinants essential for agonist and antagonist binding (3), TM2 is thought to form the inner wall of the chloride channel (4).
The glycine receptor ␣ subunit of rodent central nervous system exists in different subtypes (␣1-␣4) encoded by distinct genes (1). In the murine genome, the corresponding loci have been localized on chromosomes 11 (␣1), X (␣2, ␣4), and 8 (␣3), respectively (5)(6)(7)(8). Further diversity is achieved by alternative splicing of primary transcripts encoding the ␣1 and ␣2 subunits (9,10). In the human, highly homologous ␣1 and ␣2 subunits have been identified by cDNA cloning (11) and assigned to the chromosomal regions 5q31.3 (12,13) and Xp21. 2-22.1 (11), respectively. In both, man and mouse mutant lines, mutations of GlyR subunit genes result in hereditary motor disorders characterized by exaggerated startle responses and increased muscle tone. Pathological alleles of the Glra1 gene are associated with the murine phenotypes oscillator (spd ot ) and spasmodic (spd) (5, 14 -16). A mutant allele of Glrb has been found to underly the molecular pathology of the spastic mouse (spa), where the intronic insertion of a LINE-1 transposable element results in aberrant splicing of Glrb primary transcripts (17,18). Resembling the situation in the mouse, a variety of GLRA1 mutant alleles have been shown to cause the human neurological disorder hyperekplexia or startle disease (12,13). In contrast, mutations of the human GLRB gene in hyperekplexia have not yet been reported (19). By analogy, the gene encoding the GlyR ␣3 subunit has to be considered a candidate gene for human and murine neurological disorders.
Here we describe the cloning and characterization of two splice variants of the human ␣3 subunit and the corresponding GLRA3 gene which was mapped to the chromosomal region 4q33-q34. Functional expression in HEK 293 cells of GlyR ␣3L and GlyR ␣3K resulted in the formation of glycine-gated chloride channels that significantly differed in desensitization behavior.

EXPERIMENTAL PROCEDURES
Isolation of GlyR ␣3 cDNA Clones-A human fetal brain library (CLONTECH, Heidelberg, Germany) was used for isolation of cDNA * This work was supported by Bundesministerium fü r Bildung und Forschung, Deutsche Forschungsgemeinschaft (SFB 539) and Deutsche Krebshilfe (10 -1124-Li1). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF0177715-AF017724, AF018157, and U93917.
** To whom correspondence should be addressed. Tel.: 49-9131-85-4190; Fax: 49-9131-85-2485; E-mail: C.-M.Becker@biochem. uni-erlangen.de. 1 The abbreviations used are: GlyR, glycine receptor; TM1-TM4, transmembrane spanning regions; GLRA, designation for human glycine receptor genes; Glra, designation for murine glycine receptor genes; PCR, polymerase chain reaction; FISH, fluorescence in situ hybridization; SSCP, single strand conformation polymorphism; UTR, untranslated region; PAC, P1 artificial chromosome; kb, kilobase(s); clones. After plating of ca. 9 ϫ 10 6 plaque forming units, screening was performed under intermediate stringency conditions using a complex cDNA probe which covered the whole open reading frames of rat ␣1, ␣2, ␣3 and human ␣1 and ␣2 cDNAs. Hybridization was carried out at 52°C in 1 mM EDTA, 0.5 M NaH 2 PO 4 /Na 2 HPO 4 , 7% SDS (w/v) and 100 mg/ml denaturated salmon sperm DNA for 12 h. Filters were washed two times in 2ϫ SSC and 0.1% SDS at 52°C for 20 min. Of the positive clones, a randomly selected sample was further analyzed by Southern blot hybridization. To this end, human ␣1and ␣2-specific cDNA probes were generated. The ␣1 probe amplified by PCR corresponded to nucleotide positions Ϫ30 to 92, while an ␣2-specific probe was generated by EcoRV and PvuII restriction digest of an ␣2 cDNA clone and covered nucleotides 1162-1262 encoding the cytoplasmic loop between TM3 and TM4. Hybridization and washing conditions were as given above, except that the temperature was 65°C. Two of the resulting clones, p7 (3 kb) and p12 (1 kb), that hybridized to neither the ␣1 nor the ␣2 probe were characterized further. All other DNA manipulations were according to standard procedures (20). DNA sequencing was performed on the ABI PRISM 377 automated DNA sequencer.
Isolation and Characterization of Genomic PAC Clones-Genomic P1 artificial chromosome (PAC) clones were obtained by screening spotted filters of a human high density PAC-library (catalog no. 704) provided by the Resource Center of the German Human Genome Project (Berlin, Germany). The radiolabeled 3-kb insert of clone p7 was employed as screening probe at hybridization and washing conditions as given above (65°C). Ten genomic PAC clones obtained were subjected to both, PCR characterization using exonic primers and Southern blot hybridization to cDNA probes. Four PAC clones (ZP2.1, LLNLP704H07287Q; ZP3.2, LLNLP704P0954Q1; ZP5.1, LLNLP704L4250Q; and ZP10.1, LLNL-P704O21267Q) were further analyzed as they turned out to represent a complete GLRA3 contig. These clones served as templates for direct, automated DNA sequencing with appropriate cDNA primers.
Northern Blot Analysis-Regional expression of the GlyR ␣3 subunit was analyzed using prefabricated RNA blots of human central nervous system tissues (human brain MTN blot II; CLONTECH, Heidelberg, Germany). Two probes covering those nucleotide positions that encode the cytoplasmic loop between TM3 and TM4 were used for isoform selective detection of both mRNA variants, ␣3L and ␣3K. Among the probes used, oligonucleotide hsa3-ish-ins (CCATATCTGAGAAACGG-TAAAACTTCTCAGTGCAAAAGCTTCTGT) contains the alternatively spliced cDNA stretch of 45 bp characteristic for the GlyR ␣3L transcript variant. In contrast, GlyR ␣3K transcripts were detected employing a radiolabeled ␣3K-cDNA amplimer from clone p7, covering nucleotides 981 to 1187. The cDNA probe was radiolabeled by random primed incorporation (PrimeIt; Stratagene, Heidelberg, Germany) of [␣-32 P]ATP, while the oligonucleotide hsa3-ish-ins was 5Ј-end-labeled with polynucleotide kinase (Promega, Mannheim, Germany) using [␥-32 P]ATP. Following the manufacturer instructions, overnight hybridizations were performed using ExpressHyb hybridization solution (CLONTECH) at 68°C for oligonucleotide and 70°C for cDNA probes. Subsequently, membranes were washed three times at 32°C for 10 min, and once at 54°C for 20 min in 2ϫ SSC and 0.1% SDS. Using intensifying screens, x-ray films were exposed at Ϫ70°C for 20 h (blot A), 52 h (blot B), and 5.5 h (blot C) as depicted in Figs. 2, A-C.
FISH and Image Processing-PAC clones ZP3.2 and ZP5.1 were labeled with digoxigenin-11-dUTP using standard nick translation protocols, and FISH was performed to normal human metaphase chromosomes separately for each PAC clone (21). For FISH, 200 ng of each labeled PAC clone were combined with 10 g of human Cot1 DNA and 10 g of salmon sperm DNA in 10 l hybridization solution. Following denaturation, the DNA was allowed to pre-anneal for 30 min at 37°C. After hybridization at 37°C overnight, posthybridization washes were performed to a stringency of 0.5ϫ SSC at 60°C. Hybridized probe was detected by anti-digoxigenin fluorescein isothiocyanate (FITC) and chromosomes were counterstained with 4,6-diamidino-2-phenylindoldihydrochloride (DAPI). The experiments were analyzed by epifluorescence microscopy. Digitized images were obtained separately for DAPI and FITC with a cooled charge-coupled device camera (Photometrics, Tucson, AZ).
SSCP Screening for GLRA3 Mutant Alleles-PCR amplimers covering exonic regions and flanking intronic sequences of the GLRA3 gene were generated from peripheral leukocyte DNA and subjected to mutational analysis by SSCP screening. Amplifications (30 cycles) were carried out in a total volume of 50 l containing 25-50 ng of genomic DNA, 10 pmol of each primer, 50 mM each dNTP, 20 mM Tris-HCl, pH 8.3, 2 mM MgCl 2 , 50 mM KCl, and 1 unit of Taq polymerase with an annealing temperature of 55°C. For denaturation, 10 l of the PCR reaction were mixed with 1.5 l of 0.5 M NaOH, 10 mM EDTA, 1.5 ml of 50% sucrose, 0.1 M EDTA, 0.1% xylene cyanol, 0.1% bromphenol blue, and subsequently heated. Denaturated fragments were separated for 1.5-2 h on a 12% polyacrylamide gel at the temperatures of 10, 15, 20, and 25°C. DNA was detected by silver staining (Amersham Pharmacia Biotech, Freiburg, Germany).
Functional Expression of ␣3 Subunit Variants-To generate an appropriate expression construct, the GlyR ␣3K open reading frame was amplified from clone p7 using the PCR primers hsa3-S1-exp (ACA-GAATTCCGTATCATGGCCCACGTGA) and hsa3-A10-exp (ACAG-GATCCCCCAGAGACTTAATCTTG). Restriction sites for EcoRI and BamHI (underlined nucleotide positions) were introduced into hsa3-S1exp and hsa3-A10-exp and used for subcloning of the amplimers obtained into the multiple cloning site of the expression vector pRk5. The cDNA clone p12, corresponding to GlyR ␣3L, lacks the 5Ј end of the open reading frame. Thus, two PCRs were required for amplification of a complete GlyR ␣3 open reading frame. Nucleotide positions Ϫ8 to 1186 were amplified from clone p7 using the primers hsa3-S1-Exp and hsa3-A10 -1167 (CCTTTGCTTGTAGACATGGT). The sequence harboring the alternatively spliced part of the GlyR ␣3L cDNA was amplified employing oligonucleotides hsa3-S8 -999 (TTTTCAGCACTTCTGGAG) and hsa3-A10-exp and clone p12 as template. Both amplimers were cut at a common PvuII restriction site following nucleotide position 1006, and the fragments resulting were ligated to yield a complete GlyR ␣3L open reading frame. The GlyR ␣3L fragment obtained was subcloned into the EcoRI and BamHI restriction sites of the expression vector pRK5. Both expression constructs, differing in the 45-bp insert of GlyR ␣3L, were verified by complete sequencing. Human embryonic kidney cells (HEK-293 cells, CRL 1573; ATCC, Manassas, VA) were transfected (22) with the human GlyR ␣3K and GlyR ␣3L expression constructs.
Electrophysiological Recording-Transfected cells were viewed with an inverted microscope (Axiovert 35, Zeiss, Jena, Germany) and continuously perfused (1 ml min Ϫ1 ) at room temperature (21-25°C) with an extracellular bath solution containing: 137 mM NaCl, 5.4 mM KCl, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 11 mM EGTA, 10 mM HEPES, adjusted to pH 7.2 with NaOH. Membrane currents were obtained from cells using an EPC-9 amplifier (HEKA Elektronik, Lambrecht, Germany) linked to an Atari STE computer controlled by HEKA software. The membrane potential was clamped at Ϫ70 mV in all experiments and agonistinduced whole-cell currents were sampled at 20 Hz. Electrodes were pulled from borosilicate glass capillaries (Hilgenberg, Malsfeld, Germany) with a Zeitz DMZ Universal Puller (Zeitz Instruments, Augsburg, Germany) to yield tip resistance of 2-4 megohm. Pipettes were filled with a solution containing: 120 mM CsCl, 20 mM tetraethylammonium chloride, 1 mM CaCl 2 , 2 mM MgCl 2 , 11 mM EGTA, and 10 mM HEPES (pH 7.2). Series resistances after whole-cell formation were compensated for 50 -90%. Superfusion was performed with a DAD-12 (Adams and List, Westbury, NY) drug application system. Dose-response curves of agonist-induced peak currents were normalized to the maximum value and data were fitted with the sigmoidal Hill equation using the Levenberg-Marquardt algorithm. Results are expressed as means Ϯ standard deviation. Double exponential curves were fitted to the desensitizing phase of the GlyR responses.

RESULTS
cDNA Cloning and Characterization of GlyR ␣3 Transcripts-To identify novel variants of the ligand binding GlyR subunit from human central nervous system, a fetal brain cDNA library was screened using a complex probe composed of various human and rodent ␣ subunit cDNAs (␣1-␣3). From a random sample comprising 20 out of Ͼ100 hybridizing clones, transcripts of the human GlyR ␣1 and ␣2 subunit genes were excluded by Southern blot hybridization to ␣1 and ␣2 cDNA fragments. While none of the clones hybridized to the ␣1 fragment, 18 clones were recognized by the ␣2-specific probe. This is consistent with a prevalence of the ␣2 subunit in the human fetal central nervous system, resembling the situation in the rodent (23,24). The nonhybridizing clones p7 and p12 were found to contain inserts of 3069 and 940 base pairs, respectively. As revealed by DNA sequencing, clone p12 was highly homologous to the rat GlyR ␣3 subunit cDNA (25) representing a 3Ј partial clone that lacked a 5Ј segment of 733 nucleotides as counted from the translation start (Fig. 1). In contrast, clone p7 contained a complete open reading frame as well as large parts of 5Ј-and 3Ј-untranslated regions (UTRs). Compared with clone p12 and the rat ␣3 subunit cDNA, clone p7 lacked a stretch of 45 bp corresponding to nucleotides 1072-1116 of the coding sequence (Fig. 1), indicative of alternative splicing. This predicts a loss of 15 amino acids positioned within the cytoplasmic loop connecting TM3 and TM4 of the subunit polypeptide, hence referred to as GlyR ␣3K. Moreover, the sequences of both clones differ within the 3Ј-UTR starting with nucleotide posi- tion 1649. This would be compatible with the occurrence of two distinct transcripts resulting from a complex splice event where excision of the alternative sequence of 45 bp coincides with a splice event in the 3Ј-UTR. A search for internal homologies revealed that a 31-bp long internal repeat existed within the GlyR ␣3L transcript. The sequence of nucleotide positions 692-723 is identical to the reverse sequence comprising nucleotides 1642-1672, predicting a putative loop structure with a 31-bp stem length. In the GlyR ␣3K transcript, this motif does not exist as both variants differ starting from nucleotide position 1649. Except for the deletion present in GlyR ␣3K, the human GlyR ␣3 open reading frame was 88% identical at the nucleotide and over 98% identical at amino acid level to the rat ␣3 subunit. Of seven amino acid substitutions, one affects the signal peptide, five are localized within the cytoplasmic loop, and an additional substitution is in the C-terminal part. Thus, interspecies sequence divergency is highest in the cytoplasmic loop, reminiscent of other GlyR ␣ subunit variants (2, 6). A search for transcription regulator motifs identified a consensus sequence of the neuron-restrictive silencer element situated within the 5Ј-UTR, 145 bp in front of the translation start (Fig.  1). This recently described short DNA element contributes to the suppression of neuron-specific genes in non-neuronal tissues (26).
Regional distribution of the alternatively spliced transcripts GlyR ␣3L and GlyR ␣3K in adult human central nervous system was assessed by Northern blot hybridization. The probes used selectively recognized isoform-specific sequences coding for the divergent region within the cytoplasmic loop between TM3 and TM4. This led to the detection of mRNA species of 9 and 2.4 kb corresponding to the GlyR ␣3K and GlyR ␣3L transcripts, respectively (Fig. 2). Both subunits were widely expressed throughout most central nervous system regions examined.
The Structure and Chromosomal Localization of Human GLRA3 Gene-A human genomic PAC library (27) was screened using the radiolabeled 3-kb insert of cDNA clone p7. Based on the cDNA sequences, ten positive PAC clones were characterized further by PCR and Southern hybridization. Four clones proved to be overlapping and were used for the determination of the GLRA3 gene structure by sequencing the exon-intron boundaries and flanking intronic sequences. The coding sequence was found to be distributed over ten exons (exons 1-9; Fig. 1 and Table I), while exon 10 was completely located in the 3Ј-UTR. The sequences of the exon-intron boundaries (Table I) largely match the consensus sequences determined for mammalian splice sites (28). A comparison showed that the exon-intron anatomy of GLRA3 resembled those of the human and murine ␣1, ␣2, and ␣4 genes (8,12,29). Positions of exon-intron boundaries proved to be highly conserved, and gene structure homology was highest with GLRA1 sharing the same lengths for exons 1-8. The alternatively spliced stretch of 45 bp constitutes a separate novel exon, referred to as exon 8A. The putative splice site located in the 3Ј-UTR of the ␣3 transcripts was analyzed by sequencing the corresponding genomic region contained in PAC clone ZP5.1. As the genomic sequences obtained were completely identical to the 3Ј-UTR within clone p7, the sequence starting from nucleotide 1648 in clone p7 is thought to represent the unspliced intron 9. To exclude that the persistence of intron 9, as found in clone p7, was due to a splice artifact, the expressed sequence tags (EST) data base of the GenBank TM was searched using the 3Ј-UTR sequence of this clone. Indeed, two independent cDNA clones (AA283885, AA488804) were identified that contained the identical 3Ј-UTRs covering the exon 9/intron 9 transition sequence. The chromosomal localization of GLRA3 was determined by FISH to normal human metaphase chromosomes using the two independent PAC clones ZP3.2 and ZP5.1 detected by an FITClabeled antibody against digoxigenin. From each of the two experiments, at least 40 metaphase cells were analyzed microscopically and 15 metaphase cells with particularly extended chromosomes were selected for digital image analysis. The FITC images, revealing the probe signals, were overlaid with the corresponding images of DAPI-banded chromosomes. Both PAC clones mapped to the same chromosomal bands allowing the assignment of GLRA3 to 4q33-q34 (Fig. 3). No additional signals were found in other regions of the human genome for any of the PAC clones tested.
GLRA3 Allelic Variants and the Human Hypertonic Motor Disorder, Hyperekplexia-The human neurological disorder, hyperekplexia, is caused by mutations of the GLRA1 gene in a large number of cases. In the majority of cases, however, no association could be found with allelic variants of GLRA1 (12). We therefore investigated the role of GLRA3 as a candidate gene of hereditary hyperekplexia. DNA samples from 14 patients previously excluded to carry GLRA1 coding mutations,  were subjected to SSCP screening. In analogy to GLRA1 allelic variants, where all amino acid substitutions identified are restricted to TM2 and its flanking polypeptide segments, we focused on the corresponding exons 6, 7, and 8 (12, 30 -32). In 4 out of 14 patients, an SSCP polymorphism was identified in amplimers from exon 7. Direct sequencing revealed that this polymorphism was due to a silent C to T exchange in the wobble position of codon T292. Polymorphism frequencies did not significantly differ between a sample of normal probands and affected individuals (data not shown). Functional Expression of GlyR ␣3 Variants in HEK 293 Cells-The physiological properties of glycine receptor channels encoded by both ␣3 splice variants were analyzed by patchclamp recording from transfected HEK 293 cells. Upon expression of the GlyR ␣3L and ␣3K cDNA constructs, superfusion with glycine elicited current responses characterized by maximal membrane currents (I max ) with mean amplitudes of 3988 Ϯ 1169 pA and 3690 Ϯ 737 pA, respectively (Fig. 4A, Table II). Half-maximal responses (EC 50 ) were observed at glycine concentrations of 54 Ϯ 12 M (n ϭ 7) and 64 Ϯ 14 M (n ϭ 5) for GlyR ␣3L and GlyR ␣3K, respectively (Fig. 4B, Table IIc).
When symmetrical chloride concentrations (145 mM) were used, analysis of the I-V relationships of glycine-induced currents produced no significant differences in reversal potentials of about 0 mV (data not shown). With both variants, the glycine currents induced proved to be sensitive to strychnine, but revealed no detectable differences in the strychnine-affinity (data not shown). To unravel differential efficacies of glycinergic agonists, the relative sizes of maximum currents induced by glycine, ␤-alanine, and taurine were compared for both GlyR ␣3 splice variants. No differences in the efficacies of ␤-alanine (10 mM) and taurine (10 mM) were observed as compared with glycine (1 mM, data not shown).
Glycine receptor currents show a pronounced desensitization behavior. Using the agonist glycine at near saturating concentrations (100 M), the time courses of current desensitization were investigated for both ␣3 splice variants. As depicted in Fig. 4C for representative responses, application of glycine (10 s) to cells expressing these splice variants revealed distinct desensitization kinetics. In most of the experiments, both variants exhibited time courses of desensitization that fitted a double exponential decay, where the relative amplitude of the second component consistently accounted for less than 20% of the first one. Receptors expressed from ␣3L constructs desensitized more slowly and declined to a lesser extent than ␣3K receptor channels (Table II). DISCUSSION Here, we present the structure and chromosomal localization of the human GlyR ␣3 subunit gene (GLRA3), and its splice variants GlyR ␣3L and GlyR ␣3K. As revealed by functional expression, these variants give rise to glycine-gated chloride channels differing in desensitization kinetics.
Molecular cloning led to the identification of the human GlyR ␣3 subunit that was found to exist in two splice variants differing in sizes of the polypeptides encoded. The amino acid sequences of both variants were highly homologous to the previously characterized rodent polypeptide (25), indicating that the gene family of glycine receptor subunits is conserved during phylogeny (6,8,25). While subunit variant ␣3L represented the homologue of the rat ␣3 transcript previously described (25), a 45-bp segment was deleted from the novel transcript ␣3K. When compared with the GLRA3 gene structure, it became apparent that this alternative segment reflected a distinct exon, termed exon 8A, which codes for 15 amino acids situated within the putative cytoplasmic loop connecting TM3 and TM4 of the mature polypeptide.
Both transcripts exhibited further divergency. The 3Ј-UTR of transcript ␣3K represented a continuous copy of the corresponding genomic region. In contrast, a diverging sequence was contained in the 3Ј-UTR of ␣3L, indicative of a further splice event. The difference in ␣3 transcript sizes (2.4 versus 9 kb) as observed by Northern analysis is most likely explained by the persistence of large intronic sequences within the 3Ј-UTR of the ␣3K mRNA. This analysis also showed that the inclusion of exon 8A is linked to the excision of these large 3Ј sequences (Ͼ6 kb) from the ␣3L transcript, generating an inverted repeat of 31 bp. A sequence motif overlapping the boundary between exons 6 and 7 is invertedly repeated by a stretch of nucleotides contained within the segment of the 3Ј-UTR unique for variant ␣3L. As inverted repeats are capable to form stem structures of RNA loops, these ␣3 transcript variants are likely to differ substantially in mRNA secondary structures. Thus, inclusion of 3Ј intronic sequences may lead to an altered mRNA folding, suggesting the formation of variant-specific mRNA secondary structures. The functional importance of these mRNA structures is not understood. A search for specific RNA consensus sequences revealed, however, that a cluster of three pentanucleotides occurred in the 5Ј-UTR, reminiscent of the consensus sequence (UCAU(N) 0 -2 ) 3 recognized by the RNA binding protein Nova-1 (33). Binding of Nova-1 has been demonstrated for primary transcripts encoding the human GlyR ␣2 subunit (33), suggesting that the motif analyzed here may indeed serve a similar function. Taken together, it may be tempting to speculate that splicing regulates the formation of long stem-loop RNA structures, thus exposing determinants for RNA-protein interaction.
The GlyR ␣3 variants generated by alternative splicing differ within the cytoplasmic loop between TM3 and TM4. A similar heterogeneity exists for the splice variants of the GlyR ␣1 subunit (9) and the 5-HT 3 receptor (34), where usage of an alternate acceptor splice site for exon 9 results in the insertion of eight and six amino acids, respectively, within a homologous position. These structural characteristics of variants ␣3K and ␣3L coincide with distinct desensitization behaviors of the re- ceptor channels formed upon recombinant expression, thus defining the cytoplasmic loop as an important determinant of channel inactivation kinetics. Further analysis reveals the alternative insert to carry a consensus target sequence for casein kinase II-dependent phosphorylation of serine 370 (35). Indeed, protein phosphorylation may significantly affect channel desensitization kinetics, as studied in detail with the nicotinic acetylcholine receptor (36,37). In the ␥ and ␦ subunits of this receptor, those phosphorylation sites associated with this effect are also situated within the cytoplasmic loop between TM3 and TM4. In addition, phosphorylation of the GlyR ␣1 large cytoplasmic loop by protein kinases A and C affects the sizes of glycine-evoked membrane currents in opposing ways (38,39).
In contrast to the functional expression of GlyR ␣3 splice variants in HEK 293 cells, no physiological differences were apparent upon heterologous expression of the GlyR ␣1 and 5-HT 3 receptor splice variants in Xenopus laevis oocytes (9,34). It should be noted, however, that the phosphorylation background is high in Xenopus laevis oocytes due to a high basal level of protein kinase A activity, thus potentially masking functional differences (37). It remains to be shown whether the differences in GlyR ␣3 desensitization kinetics can be attributed to changes in phosphorylation status or to alterations in receptor architecture due to inclusion of the additional peptide sequence.
Analysis of the GLRA3 gene structure revealed a high degree of homology to the previously characterized glycine receptor subunit genes in mouse and man (8,12,19,29). The alternatively spliced exon 8A, however, represents a structural element unique among glycine receptor genes. Mapping of the GLRA3 gene locus revealed its chromosomal localization in the vicinity of the GLRB gene, which was assigned to the human chromosomal band 4q31.3 (19,40). This chromosomal region is linked by synteny homology to a region on mouse chromosome 8, where the murine Glra3 gene is situated (7). While no obvious correlations to currently known disease loci exist, the hu- ␣3L cDNA (f). For both cDNAs, data were normalized to peak amplitudes (I max ) obtained at saturating concentrations of glycine; ␣3K subunit I max ϭ 3690 pA, EC 50 ϭ 64 Ϯ 14 M; ␣3L subunit I max ϭ 3988 pA, EC 50 ϭ 54 Ϯ 12 M. In both cases, Hill coefficients were above 2. C, desensitization of responses to 100 M glycine of the two different splice forms of the GlyR ␣3. Receptors in this example composed of ␣3L desensitized more slowly ( 1 ϭ 5.8 s, upper trace) than receptors composed of ␣3K subunits ( 1 ϭ 630 ms, lower trace) and declined to a lesser extent (47% versus 85%). The traces show in detail the first 5 s of desensitization. Double exponential curves were fitted to the desensitizing phase of the GlyR responses.