A Novel Glycine Receptor Subunit Splice Variant Predicts an Unorthodox Transmembrane Topology ASSEMBLY INTO HETEROMERIC RECEPTOR COMPLEXES *

Jana Oertel, Carmen Villmann, Helmut Kettenmann, Frank Kirchhoff , and Cord-Michael Becker From the Institut für Biochemie, Emil-Fischer-Zentrum, Universität Erlangen-Nürnberg, Fahrstrasse 17, 91054 Erlangen, Germany, the Max-Delbrück-Centrum für Molekulare Medizin, Robert-Rössle-Strasse 10, 13092 Berlin, Germany, and the Max-Planck Institut für Experimentelle Medizin, Hermann-Rein-Strasse 3, 37075 Göttingen, Germany

Glycine receptors (GlyRs) 2 belong to the superfamily of Cysloop receptors, which also include nicotinic acetylcholine receptors, ionotropic ␥-aminobutyric acid receptors (GABA A and GABA C ), and the ionotropic serotonin receptor subtype 5HT 3 (1).GlyRs are pentameric assemblies of five subunits surrounding a central ion-conducting pore that mediates rapid inhibitory neurotransmission in the spinal cord and brainstem (2,3).Common features of all Cys-loop receptor subunits include four transmembrane regions (TM1-TM4), where TM2 forms the ion channel pore, and a large extracellular N-terminal ligand-binding domain.This domain shows significant structural homology to the acetylcholine-binding protein of Lymnaea stagnalis (4), a member of the immunoglobulin-like superfamily of proteins.By association of ␤ subunits with the intracellular anchor protein gephyrin, GlyRs are clustered at the postsynaptic membrane.The gephyrin binding motif of the ␤ subunit polypeptide has been mapped to the long intracellular TM3-4 loop (5)(6)(7)(8)(9).
The GlyR ␤ subunit gene is widely transcribed throughout the central nervous system of neonatal and adult rodents.Starting at embryonic day 14, ␤ transcripts are first detectable by in situ hybridization in rat spinal cord and telencephalon (10,11).Expression of the ligand binding ␣ subunit variants (␣1-␣4) is highly regulated during development (12,13).Further GlyR subunit heterogeneity is generated by alternative splicing and RNA editing (␣1 and ␣1ins; ␣2-3A and -3B; ␣3L and ␣3K) (11,14,15).Whereas, there are no reports about alternative splicing of the ␤ subunit, aberrant splicing of GlyR ␤ subunit transcripts underlies neurological disorders in a mouse mutant and human hyperekplexia (16).The pathological phenotype of the recessive mouse mutant spastic, where a LINE1 element is inserted into intron 5 of the Glrb gene results in exon skipping and a dramatic reduction in GlyR number (17,18).Likewise, compound heterozygote mutations of the human ␤ subunit gene GLRB, with one allele resulting in skipping of exon 5, cause the human neurological disorder hyperekplexia (16).
As evident from biochemical cross-linking experiments (3) and recombinant ion channel physiology (19) pentameric GlyRs display a subunit stoichiometry of ␣ 2 ␤ 3 .Upon heterologous expression, however, GlyR ␣ subunits are able to assemble into functional homopentameric channels in Xenopus oocytes or mammalian cells (10,12,20,21).When co-expressed with ␣ subunits, ␤ subunits incorporate into heteromeric GlyRs and contribute to intracellular transport during protein biogenesis and receptor anchoring.Moreover, the ␤ subunit modifies physiological and pharmacological properties of GlyR chloride channels as demonstrated by altered single channel conductance and induction of picrotoxin insensitivity of heteromeric ␣1/␤ as compared with homomeric ␣1 receptors (22)(23)(24).As a consequence of the interaction of the ␤ subunit with gephyrin, GlyRs are clustered at the postsynaptic membrane.Gephyrin is highly enriched at the cytoplasmic face of the inhibitory postsynaptic membrane differentiations and colocalizes with both GlyRs and some GABA A receptor variants in spinal cord, retina, and other brain regions (25)(26)(27)(28)(29).
In this study, we identified by single cell RT-PCR a novel GlyR ␤ subunit splice variant that lacks exon 7.This variant, termed ␤⌬7, is highly expressed in glial cells of rat spinal cord and in cultured astrocytes from mouse brain.The variant ␤⌬7 was also detectable in total RNA preparations from mouse spinal cord, cortex, liver, and heart.Exclusion of exon 7 predicts an altered transmembrane topology of the ␤ polypeptide with TM1 and TM2 missing.Recombinant expression revealed that ␤⌬7 forms receptor complexes with the ligand binding ␣1 subunit as well as the anchor protein gephyrin.

EXPERIMENTAL PROCEDURES
Single Cell RT-PCR, Total RNA Extraction, and cDNA Synthesis and Amplification-Harvesting of cytoplasmic RNA, reverse transcription, and PCR were done as described (30,31).In brief, patch clamp recordings were done and cellular RNA was harvested by applying negative pressure to the patch pipette filled with internal solution under visual control.Care was taken to avoid contaminations by material from adjacent cells.Immediately after retraction of the pipette, its content was expelled into a sterilized PCR tube containing 0.5 l of 10 mM dNTPs (Roche), 1 l of 50 M random hexamer primers, 1 l of 0.1 M dithiothreitol, and 0.5 l of 30 units/ml RNasin (Promega).Reverse transcription was started by adding 0.5 l of 200 units/l of reverse transcriptase (Superscript II, Invitrogen) to a final volume of 10 l.After incubation for 1 h at 37 °C, reaction was stopped by freezing at Ϫ20 °C.
The polymerase chain reaction was performed in a final volume of 50 l containing 10 l of reverse transcription reaction, 10 pmol, each, of sense and antisense primer, 5 l of ϫ10 Taq polymerase buffer, and 2.5 units of Taq polymerase (Stratagene).A PerkinElmer Thermocycler 9600 was used set to 5 min at 94 °C for primary denaturation, 35-40 cycles for amplification (94 °C denaturation, 30 s; 45-51 °C annealing, 30 s; 72 °C elongation, 30 s (ϩ1 s per cycle), and 10 min at 72 °C for elongation.After the final elongation step, PCR tubes were cooled to 4 °C and analyzed or frozen for later use.For detection of GlyR␤-specific cDNA sequences, two PCR were performed using different ␤ specific primer pairs, bS1 and bAS1 in the first round, and "nested" primers (bS2/bAS2 generating DNA fragments of 509 bp length) in the second round.Primer sequences were bS1/bS2 (5Ј-GCTGCAGAGGACCTTGCCCGTGTGC-3Ј), bAS1 (5Ј-GCTCGAGCCACACATCCAGTGCCTT-3Ј), and bAS2 (5Ј-GTCGTCTTCGAACGTAGGACACTTTGGG-3Ј) (10).Due to interspersed intronic sequences, amplification of contaminating genomic DNA could be distinguished from amplified ␤ cDNA fragments by the size differences of the respective amplimers.As a positive control, cDNA derived from spinal cord total RNA was used instead of cytoplasmic RNA from individual cells.For negative control, either buffer (internal pipette solution) was used instead of harvested RNA or cytoplasmic RNA obtained from glial cell displaying no glycine-evoked inward currents (n ϭ 3).For sequencing, PCR amplification products were purified and cloned into pBluescript (Stratagene).
Transfection of HEK293 and COS7 Cells-COS7 cells were transfected (10 g of plasmid/10-cm dish) using DEAE-dextran (10 mg/ml).HEK293 cells were transiently transfected (10 g of plasmid/10-cm dish) using a modified calcium-phosphate precipitation method (32).Green fluorescent protein was cotransfected as a control for transfection efficiency.All experiments for immunocytochemistry, protein biochemistry, and electrophysiological recordings were performed 48 h following transfection.For glycosylation analysis, tunicamycin (5 g/ml) was added 5 h after transfection, and the cells were incubated in the presence of tunicamycin for an additional 24 h prior cell lysis.
Electrophysiological Analysis of Spinal Cord Glia and Transfected HEK293 Cells-Preparation of spinal cord slices and electrophysiological setup were done as described previously (30).For recordings, cells were selected that were located 10 -30 m below the surface of the slice and displayed a clear, dark membrane surface.Membrane currents were measured applying the patch clamp technique in a whole cell recording configuration (33).During application of a series of de-and hyperpolarizing voltage steps (50 ms, ranging from Ϫ160 to ϩ20 mV in 10-mV increments), whole cell membrane currents were recorded and cells were classified as astrocytes, oligodendrocytes, or glial progenitor cells/NG2 glia (30,34).Current signals were amplified with conventional electronics (EPC-7 or EPC-9 amplifier, HEKA), filtered at 3 kHz, and sampled at 5 kHz.A standard bathing solution was used in the experiments with the following composition: 134.0 mM NaCl, 2.5 mM KCl, 2.0 mM CaCl 2 , 1. Biotinylation of Cell Surface Protein and in Vitro Deglycosylation-The biotinylation assay was performed 48 h posttransfection of COS7 cells grown in 6-cm dishes as described previously with little variations (35).In brief, after labeling of surface protein with EZ-link TM sulfo-(NHS)-S-S biotin (Pierce) and cell lysis, supernatants were incubated with 50 l of streptavidin-Sepharose beads (Sigma) for 4 h at 4 °C while rotating.Biotinylated proteins were eluted by boiling the probes in 1ϫ sample buffer for 5 min.For Endo H digest, biotinylated surface proteins were denatured in 1ϫ denaturing buffer at 100 °C for 10 min, Endo H (New England Biolabs) digest was done following the manufacturer's protocol.
Antibody Generation-A peptide epitope bridging the border of exons 6 and 8 within the ␤⌬7 sequence (TKYYK-GTGIFSVLSLASECEC) was chosen based on antigenicity prediction.The peptide was synthesized using Fmoc (N-(9fluorenyl)methoxycarbonyl)-PAL-PEG-PS-resin to construct a peptide-amide (9050 PlusPep Synthesizer, Millipore).The integrity and purity of the lyophilized crude peptide were verified by reversed-phase high performance liquid chromatography analysis and mass-assisted laser desorption ionization time of flight mass spectrometry.The crude peptide was coupled to keyhole limpet hemocyanin via the heterobifunctional crosslinker sulfo-succinimidyl 4-N-maleimidomethyl cyclohexane 1-carboxylate.Rabbits (chinchilla bastards) were immunized with 400 g of peptide conjugate at days 0, 14, 28, and 40.The first test serum was taken 6 weeks after the first boost.Preadsorbed ␤⌬7 antibody K12 was generated by overnight coincubation of serum (dilution 1:20) with a membrane preparation from HEK293 cells transfected with the ␤ subunit.
Lysate and Membrane Preparation-For total cell lysates, transfected cells were lysed in 25 mM potassium phosphate buffer, pH 7.4, containing 5 mM EDTA and a protease inhibitor mixture (Roche).Membranes were separated from nuclei by ultracentrifugation (140,000 ϫ g for 1 h).Pellets were resuspended in reducing sample buffer (2% SDS, 5% ␤-mercaptoethanol in 0.68 M Tris, pH 6.8) and separated on a 10% polyacrylamide gel (20 l of lysate/lane).For membrane protein analysis, crude cell membranes were prepared from transfected cells and mouse tissues as described previously (36).For detection of myc-tagged ␤ or ␤⌬7 the monoclonal and polyclonal anti-myc antibodies 9E10 (Santa Cruz) were used.The native ␤ subunits were detected with the polyclonal anti-GlyR␤-(H170) antibody (Santa Cruz).For detection of the ␣ subunits mAb4a hybridoma supernatant was used, for detection of gephyrin a monoclonal antigephyrin antibody (Clontech).Proteins were visualized using either horseradish peroxidase-conjugated goat antimouse IgG (Dianova) followed by detection using the ECLPlus system (GE Healthcare) or goat anti-rabbit IgG-Cy5 (Dianova) analyzed with fluoroimager STORM 860 (GE Healthcare).
Immunoprecipitation-Crude cell membranes (1 mg) were resuspended in 20 mM Tris-HCl, pH 7.0, and sodium desoxycholate (10% in 50 mM Tris-HCl, pH 9.0), and incubated for 30 min at 30 °C, followed by solubilization with 2% Triton X-100 in 50 mM Tris-HCl, pH 7.0, while gently agitating for an additional 30 min at 30 °C.After centrifugation (100,000 ϫ g, 30 min), protein complexes were precipitated using polyclonal anti-myc antibody (1:100) by rotating the supernatants overnight at 4 °C.Prior to precipitation, a supernatant sample was harvested for further analysis.Protein complexes were incubated with protein A-conjugated Sepharose beads (Sigma).Following a 1-h incubation at 4 °C and three washing steps (20 mM Tris-HCl, pH 7.0, 0.16 M NaCl), the immunosorbent was collected at 3,000 ϫ g.The precipitated proteins were eluted by heating at 95 °C for 5 min in 2ϫ SDS sample buffer.
Topology Prediction-The topology of GlyR ␤ and ␤⌬7 was predicted according to Expasy for post-translational modification.The prediction of transmembrane domains was based on a hidden Markov model (37,38) and the predict protein server (39).

Single Cell RT-PCR Analysis of GlyR Expression in Rat Spinal
Cord Glial Cells-The starting point of this study was the analysis of GlyR subunit transcripts in glial cells, as investigated in spinal cord slices of young rats (1-3 weeks old).The majority of cells were astrocytes as they showed passive membrane properties as revealed by their whole cell membrane currents (Fig. 1A, G2) (30,34).A second population of cells exhibited voltagedependent currents characteristic for glial progenitor or NG2 glial cells (Fig. 1A, P1).Because in some cells we also observed a small component of symmetrical decaying in-and outward potassium currents, we cannot exclude the inclusion of some oligodendrocytes in our study (not shown).Glycine-evoked currents were recorded by the patch clamp technique in the whole cell configuration (Fig. 1, A and B).Thereafter, cytoplasmic RNA was harvested by applying negative pressure to the patch pipette.Surprisingly, subsequent RT-PCR analysis using nested PCR with GlyR ␤-specific primers detected the expression of cDNA fragments of 356 bp, instead of 509 bp as expected from the known GlyR ␤ subunit cDNA sequence (Fig. 1, C and  D). cDNA sequencing of the subcloned fragments of 356 bp revealed a novel alternatively spliced variant of the GlyR ␤ subunit lacking exon 7.This novel variant is designated GlyR ␤⌬7.
Deleted from the ␤⌬7 transcript, exon 7 encodes a region ranging from the C-terminal part of the large extracellular domain, TM1 and the intracellular loop between TM1 and TM2 to the first four amino acid residues corresponding to TM2 (Fig. 2A).Analysis of the polypeptide variant encoded by the ␤⌬7 transcript predicted a primary structure where TM1 and the loop connecting TM1 and TM2 were missing, suggesting a dramatic change in subunit topology (Fig. 2, B and C).Moreover, four positions of the amino acids corresponding to the pore-forming TM2 were also missing.Based on hydrophobicity analysis, only two TMs were postulated for ␤⌬7, suggesting a loss of TM2 (Fig. 2C, middle panel).Rather, prediction of secondary structure (39) suggested that the remaining amino acid residues of the former TM2 substitute for the 10th ␤-sheet of the N-terminal domain that are lost by exon 7 deletion.Due to loss of exon 7, the secondary structure probability of the former TM2 positions changes from ␣-helical to ␤-strand, completing the fold of the acetylcholine-binding protein-like N-terminal domain (Fig. 2, B and C, right panels) (4,19).Concurrently, the ion channel pore characteristic of the Cys-loop receptor subunits is lost in ␤⌬7.
To functionally characterize the ␤⌬7 polypeptide, we cloned the coding sequence into the eukaryotic expression vector pRK7.In addition, ␤ and ␤⌬7 constructs were generated carrying an N-terminal c-myc epitope (9E10) recognized by specific antibodies (Fig. 2A).
Cell Type and Tissue Distribution of ␤ Subunit Transcript Variants in Situ-Analysis of the tissue distribution of the major ␤ subunit transcript variants was extended to spinal cord (sc), cerebral cortex (cx), heart (he), and liver (li) of adult mice (Fig. 3A).Cortex contained high levels of the novel ␤⌬7 splice variant, whereas liver and heart yielded a weak signal for this transcript.In adult spinal cord the signal for the read-through ␤ transcript was barely detectable.Analysis of ␤⌬7 transcript distribution in embryonic, neonatal, and adult central nervous system revealed no significant developmental regulation (not shown).
Consistent with the single cell PCR data, RT-PCR analysis of ␤ subunit transcripts present in primary astrocytes and neurons cultured from the mouse central nervous system confirmed the presence of the two major ␤ subunit transcripts, corresponding to the read-through ␤ transcript (606 bp) and the variant lacking exon 7 (453 bp) (Fig. 3B).In cultured astrocytes, the ␤⌬7 transcript prevailed, whereas the read-through ␤ variant predominated in spinal cord neurons.
In Vivo Expression of ␤⌬7-To verify if the ␤⌬7 transcript is translated, membrane preparations from neuronal and nonneuronal tissues (cortex, spinal cord, and liver) were analyzed for concomitance of the read-through ␤ subunit and the splice variant ␤⌬7 (Fig. 4).Using the commercial antibody H170 directed against GlyR ␤ variants, we indeed detected an antigen of 53 kDa molecular mass in cortex, spinal cord, and liver corresponding to ␤⌬7 (Fig. 4A).The staining intensity of the presumptive ␤⌬7 was not significantly different from the fulllength ␤ antigen.As evident from an Endo H digest (Fig. 4A, right panel), the observed double bands do not represent glycosylated and unglycosylated forms of the full-length ␤ variant.After the Endo H digest, both bands were still present, running in a slightly lower position than those of the non-digested preparation.For immunological discrimination of ␤⌬7 from the A, a series of de-and hyperpolarizing voltage steps (Ϫ160 to 20 mV) from a holding potential of Ϫ70 mV was applied and membrane currents were recorded.The delayed rectifying outward current activated at the depolarizing potential identifies cell number 1 as a precursor cell (P).The symmetrical pattern of in-and outward currents of cell number 2 reflects the current pattern of a mature glial cell (G).B, glycine (1 mM) applied while clamping the membrane potential at Ϫ70 mV evoked a whole cell inward current.C, primer selection for amplification of GlyR ␤ transcripts.For the nested PCR approach two different primer pairs were selected from exons 2, 4, and 8. Using the primer pair bS2 and bAS2 a cDNA fragment of 509 bp for the full-length GlyR ␤ is obtained, whereas cDNAs encoding GlyR␤⌬7 give rise to a truncated fragment of 356 bp lacking the sequence region of exon 7. D, amplified cDNA fragments obtained from single cell RT-PCR of the precursor cell (P) and two mature glial cells (G2 and G3) revealed a length of 356 bp employing GlyR ␤ subunit-specific primers.Lane 4 reflects the negative control of cytoplasmatic RNA obtained from glial cells displaying no glycine-evoked inward current.
read-through ␤ variant, we generated a polyclonal antibody (K12) against the bridging epitope between exons 6 and 8 that is characteristic of the deletion variant.In membranes from transfected HEK cells, the ␤⌬7-specific antibody K12 recognized only the splice variant, but not the full-length ␤ subunit (Fig. 4B).In membranes from mouse liver, a double band was detected, most likely representing additional heterogeneity, e.g. the presence of glycosylated and unglycosylated isoforms of ␤⌬7, or additional splice variants.Preadsorption of K12 against full-length ␤ subunit antigen did not detectably alter the staining pattern, arguing for the specificity of K12 antiserum (Fig. 4B).
Cell Surface Expression of the Variants GlyR ␤ and ␤⌬7-The topological model of ␤⌬7 predicted a deletion of structural elements that are essential for proper transmembrane folding of Cys-loop receptors of the nicotinic acetylcholine receptor superfamily (19).Thus, we asked whether the ␤⌬7 polypeptide was efficiently integrated into the plasma membrane.To this end, ␤⌬7 polypeptide was recombinantly expressed in the presence and absence of ␣1 subunits, and its subcellular distribution was determined by comparing intracellular and surface protein fractions (Fig. 5A).Surface proteins were labeled by biotinylation using sulfo-NHS-S-S-coupled biotin.Consistent with earlier reports, the ␣1 subunit of GlyR, detectable by monoclonal antibody mAb4a, was detected in both the intracellular and the cell surface fraction (Fig. 5B).The cytoplas-matic anchor protein of the GlyR complex, gephyrin, served as a control for the exclusion of cytoplasmatic protein from extracellular protein biotinylation (Fig. 5B).
Previous studies (40,41) showed that an efficient plasma membrane integration of the GlyR ␤ subunit requires a complex formation with ligand binding ␣ subunits.Surprisingly, our experiments yielded a strong surface and intracellular expression signal for both, ␤ (60 kDa) and ␤⌬7 (55 kDa) constructs, even when expressed alone (Fig. 5A).For each of the constructs, however, two bands of distinct molecular weights appeared.As the amino acid sequence of the ␤ subunit harbors two extracellular glycosylation sites, this heterogeneity may be attributed to glycosylation (␤, 56 kDa, and ␤⌬7, 51 kDa).Note that the smaller ␤ subunit antigen of both variants is expressed at the cell surface in COS7 cells.Coexpression of both ␤ variants with the ␣1 subunit did not significantly increase the relative abundance of the biotinylated surface fraction of the ␤ and ␤⌬7 polypeptides (Fig. 5A).
Integration of the subunit variants ␤ and ␤⌬7 into the plasma membrane of COS7 cells was also followed by immunocytochemistry (Fig. 5, C and D).As seen in permeabilized cells (Fig. 5C), the fraction of protein accumulating in intracellular compartments exceeded that of surface ␤ protein (Fig. 5D).The subcellular distribution of tagged ␤ antigens suggested that a large fraction of the intracellular ␤ and ␤⌬7 variants were retained in the endoplasmic reticulum.Indeed, co-localization with endoplasmic reticulum-specific proteins was confirmed by co-expression with DsRed-endoplasmic reticulum (not shown).
Glycosylation Studies on GlyR ␤ and ␤⌬7-The extracellular domain carries two potential N-glycosylation consensus sequences (NX(S/T): NST 54 -56 and NCT 242-244 ), which may account for the polypeptide heterogeneity observed upon recombinant expression.Transfected COS7 cells were grown in the presence of tunicamycin, an inhibitor of formation of the core oligosaccharide.In tunicamycin-treated cells, only the low molecular weight species was present, indicating the unglycosylated forms of the ␤ and ␤⌬7 polypeptides (Fig. 6A).Conversely, the larger molecular weight bands represent the glycosylated ␤ and ␤⌬7 polypeptides.Glycosylation could also be demonstrated by sensitivity of the glycosylated polypeptides to Endo H. Indeed, when surface protein fractions were digested with Endo H, only the smaller species of ␤⌬7 was detected, consistent with an unglycosylated structure of the smaller polypeptide (Fig. 6B).
Association of ␤⌬7 with the ␣1 Polypeptide and Gephyrin-Clustering of GlyRs at the postsynaptic membrane depends on an interaction of a motif within the TM3-4 loop of the ␤ subunit with the cytoplasmatic anchor protein gephyrin (8).For formation of molecular complexes in transfected HEK293 cells, ␤⌬7 polypeptide was analyzed in immunoprecipitation studies upon recombinant co-expression with the ␣1 subunit and gephyrin (ratio ␣1/␤ or ␤⌬7/gephyrin: 1/5/5).Using antibodies against the myc epitope of ␤ or ␤⌬7 constructs, significant amounts of GlyR␣1 polypeptide (Fig. 6C) and gephyrin (Fig. 6D) were co-immunoprecipitated.These data indicated that the ␤⌬7 polypeptide, despite its fundamentally altered transmembrane topology, is able to form molecular protein complexes with both, the ␣1 subunit and gephyrin.
Is There a Physiological Response of ␤⌬7?-Upon inclusion of the ␤ subunit, heteromeric ␣1␤ chloride channels acquire a relative resistance to the plant toxin picrotoxin, which is absent from ␣1 homomers (23).Indeed, HEK293 cells transfected with ␣1 subunit cDNA yielded glycine-induced chloride currents sensitive to picrotoxin (19 Ϯ 5% I max ), consistent with the formation of ␣1 homomers (Fig. 7, A and D).Despite the presence of picrotoxin, glycine currents were largely reconstituted upon co-transfection of the full-length ␤ subunit (58 Ϯ 11% I max , Fig. 7, B and D).However, ␣1␤⌬7 heteromers were efficiently blocked by picrotoxin (10 Ϯ 2.4% I max ), indistinguishable from the situation in ␣1 homomeric receptor channels (Fig. 7, C and  D).The glycine-gated maximal currents (I max ) of both heteromeric receptor complexes ␣1␤ and ␣1␤⌬7, however, were A, cortex (cx), spinal cord (sc), heart (he), and liver (li) were used for ␤ and ␤⌬7 transcript amplification.␤-Actin (lower panel) was used as housekeeping gene, and reveals that the same amount of RNA was used for each probe.B, transcript level of ␤⌬7 in cultured astrocytes compared with spinal cord neurons.␤ and ␤⌬7 were amplified from plasmids (expected size of 606 and 453 bp) as control for the correct transcript size.Note the high amount of ␤⌬7 transcript observed in cultured glial cells at DIV 14 (cultured at P0). Positive controls: sc stained with monoclonal anti-gephyrin antibody for recognizing the 93-kDa anchor protein gephyrin (first lane) and with mAb4a to recognize the GlyR ␣1 subunit (second lane).Spinal cord, cortex, and liver were stained using the polyclonal anti-GlyR-␤-H170 antibody, which recognizes an epitope in the intracellular TM3-4 loop of the GlyR ␤ subunit.Cortex preparation was digested with Endo H.The black arrow points to the 58-kDa band of the ␤ subunit; gray arrow to the 53-kDa signal of ␤⌬7.The positive control for Endo H digest is shown in Fig. 6B.B, antibody K12 specific for the splice variant ␤⌬7 recognizes an epitope bridging the exon 6/8 border.Membrane preparation from ␤and ␤⌬7-transfected HEK cells, both constructs were stained with anti-myc antibody or K12 (1:100).K12 was tested on liver (li) and mock-transfected HEK cells compared to preadsorbed K12 antiserum.
reduced compared with ␣1 homomers (Fig. 7E).Although the formation of ␣1␤⌬7 molecular protein complexes was clearly demonstrated, the novel subunit ␤⌬7 did not confer the important pharmacological property of picrotoxin resistance to the ion channel protein.These data suggest that the ion channel pore is homomeric and formed only by ␣1 subunits.This obser-vation is consistent with the structural prediction that the deletion of exon 7 leads to the loss of a functional pore domain in ␤⌬7.

DISCUSSION
Here, we show that the rodent GlyR ␤ subunit is alternatively spliced in neural and extraneural tissues.Compared with the read-through ␤ subunit variant prevalent in adult spinal cord and cortex, the novel splice variant ␤⌬7 lacks exon 7, equivalent to a deletion of 51 amino acids.In the full-length subunit polypeptide, the deleted positions cover a segment of the N-terminal domain preceding membrane entry as well as the complete transmembrane domain TM1, the intracellular loop connecting TM1 and TM2, and the first four amino acids of the pore forming TM2.
Whereas functional GlyRs are formed by recombinant ␣ subunit expression, homomeric expression of ␤ subunits has not been seen to result in the efficient formation of glycine-gated anion channels (10).As evident from in situ hybridization (11), GlyR ␣ and ␤ subunit mRNAs show a disparate distribution in the rodent central nervous system, were ␣ subunit transcripts predominate in spinal cord and brainstem.In contrast, ␤ subunit mRNA is highly expressed throughout the central nervous system, including those forebrain regions that lack a significant ␣ subunit transcription (11).Whereas the role of the regionally isolated expression of the ␤ subunit is not understood, increasing evidence indicates that this subunit also occurs in non-neuronal cells, including astroglia, oligodendrocytes, and glial progenitor cells/NG2-glia (30).
At the transcript level, the subunit variant ␤⌬7 was highly expressed in glial cells and non-neuronal tissues, but less abun-FIGURE 5. Protein expression of ␤⌬7 in transfected COS7 cells.A, both variants were expressed alone and in combination with the ␣1-subunit in COS7 cells (amount of DNA used: 5␤:1␣).Surface expression was analyzed by labeling membrane-embedded proteins with NHS-SS-biotin.Biotinylated proteins were bound to streptavidin-agarose beads followed by separation on 10% SDS gels.The amount of surface protein (S) was compared with intracellular non-labeled protein (IC) using the monoclonal anti-myc antibody, which recognizes the myc-tagged ␤ subunits of the GlyR.The black arrow points to the 60-kDa band of ␤ and 55 kDa of ␤⌬7.An additional lower band (gray arrow) represents the unglycosylated form of both ␤ proteins (verified by Endo H digest, see Fig. 6B).␣1 and gephyrin-transfected cells were used as controls (B), ␣1 shows that a high degree of surface integration, gephyrin (93 kDa), was found only in the intracellular probe.Green fluorescent proteintransfected COS7 cells were used as negative control.C and D, subcellular localization of ␤ and ␤⌬7 in transfected COS7 cells after fixation with 4% paraformaldehyde ϩ 4% sucrose.Cells were incubated with (permeabilized) (C) and without (non-permeabilized) Triton X-100 (D).Variants ␤ and ␤⌬7 are highly expressed intracellular in the cytoplasm, but also at the cell surface.E, green fluorescent protein (GFP)-transfected cells and untransfected cells, stained only with secondary antibody.Scale bar represents 20 m.FIGURE 6.Both GlyR ␤ variants are glycosylated and form complexes with the ␣1 subunit and gephyrin.A, cell lysates from ␤and ␤⌬7-transfected COS7 cells were used for N-linked glycosylation analysis (ϪT, without tunicamycin).ϩT, treatment of cells with tunicamycin (5 g/ml) for 24 h (added to the cells 5 h post-transfection.B, IC, intracellular fraction of ␤⌬7; S, surface protein labeled with biotin and detected using anti-myc 9E10 antibody.The surface fraction of ␤⌬7-transfected cells was digested with Endo H. C and D, protein complex formation of ␣1 and ␤⌬7 polypeptides with the cytosolic anchor protein gephyrin.Membrane preparations from transfected HEK293 cells were used for immunoprecipitation. ␤ and ␤⌬7 were precipitated using the polyclonal myc antibody (1:100) and stained (C) with mAb4a for GlyR␣ subunits, solubilized protein (SL), supernatant after precipitation (SN), and precipitated fraction (IP).The black arrow points to the 48-kDa ␣1 subunit.D, protein complex formation of ␤⌬7 with the GlyR anchor protein gephyrin (dilution of antibody 1:200).The gray arrow marks the 93-kDa gephyrin.
dant in neurons.Among the tissues transcribing the Glrb gene, the variant ␤⌬7 was found in a subset of cells and tissues.Astrocytes were the most abundant cell type, playing a fundamental role in maintenance of the extracellular microenvironment during neuronal activity (42).The splice variant ␤⌬7 was highly transcribed in cerebral cortex and extraneural tissues like heart and liver, but hardly detectable in spinal cord.Whereas splicing of ␤ subunit pre-mRNA into the ␤ and ␤⌬7 variant was tissue selective, no evidence was obtained for a developmental regulation.This situation clearly differs from the strict developmental regulation of GlyR ␣ subunit variants in the central nervous system (11,41,43).Spatial expression patterns of central nervous system proteins may not only be attributed to tissue-specific transcription, but also result from a tissuespecific splicing.The neurooncological ventral antigen-1 (Nova-1) is a regulator of neuronal pre-mRNA splicing that contributes to the regulation of GlyR ␣2 variants in immature neurons (44).Similar processes may also apply to the regional regulation of ␤ pre-mRNAs.
Sequence analysis of the GlyR splice variant ␤⌬7 predicted a non-canonical topology of this Cys-loop receptor subunit.Rather than four transmembrane regions, sequence analysis of the ␤⌬7 variant predicted two membrane spanning segments.In contrast, the two Cys-loops located in the extracellular N-terminal domain that are characteristic of the GlyR family (4), are conserved in the splice variant ␤⌬7.This fundamental alteration in predicted transmembrane topology gives rise to the question whether the ␤⌬7 variant is translated into a functional polypeptide that undergoes membrane integration and assembles into functional GlyR ion channel complexes.As detectable in the mouse central nervous system in vivo, presence of the ␤⌬7 transcript variant correlated with a distinct ␤ subunit antigen of slightly lower molecular weight, suggesting that this splice variant is stably translated in cerebral cortex and extraneural tissues.
Hetero-oligomeric GlyRs are ion channel protein complexes with a subunit stoichiometry of ␣ 2 ␤ 3 , where the ␤ subunit contributes to the formation of the ligand binding site at its interface to the ␣ subunit (19).In Cys-loop receptors, neurotransmitter binding is believed to require discontinuous segments of the N-terminal extracellular regions of two adjacent subunits, situated in a distance of ϳ30 Å from the membrane (45).The amino acid positions proposed to delineate the ligand binding site are not affected by the deletion characteristic of ␤⌬7.Secondary structure prediction suggests that the amino acids remaining from the former ␣-helix of TM2 structurally replace those amino acids forming the 10th ␤-sheet of the N-terminal domain of the full-length ␤ peptide.Similar changes from an ␣-helix into a ␤-sheet stretch are also known from other proteins, e.g.␤-amyloid or prion proteins (46).The integration of the amino acid residues remaining from TM2 into the secondary structure of the N-terminal domain implies the loss of the ion channel domain.
Both, the novel subunit variant ␤⌬7, as well as the readthrough ␤ polypeptide, were sorted into the plasma membrane of COS7 cells, even in the absence of the ligand binding ␣1 subunit.Consistent with a stable expression of the ␤⌬7 polypeptide, its assembly properties were indistinguishable from those of the full-length GlyR ␤ subunit.In mature neurons, the read-through ␤ subunit polypeptide interacts with both, the ligand binding GlyR ␣ subunit and the postsynaptic anchor protein, gephyrin.The latter protein plays an important role for anchoring of both, glycine and GABA A receptors (8,47,48).By a sorting process mediated by gephyrin, ␣/␤ heterooligomeric receptors are targeted to the postsynaptic membrane (49).Likewise, recombinant ␤⌬7 polypeptide formed a stable molecular complex with both, the ␣1 subunit polypeptide and gephyrin, as evident from co-immunoprecipitation with a tagged ␤⌬7 construct.This interaction with the cytosolic protein gephyrin also suggested that the TM3-4 loop was localized intracellularly.The deletion of exon 7 does not significantly affect the assembly of hetero-oligomeric GlyR protein complexes, consistent with the identification of "assembly boxes" (40), i.e. amino acid motives located in the highly conserved N-terminal regions that are not affected by the ␤⌬7 dele- tion.Still, the quaternary structure of ␣/␤⌬7 complexes remains to be elucidated.
Whereas the post-transcriptional modification of ␤ transcripts enhances the diversity of GlyR ␤ subunit variants, the functional role of the ␤⌬7 polypeptide remains unknown.In contrast to the read-through ␤ subunit, which confers resistance to picrotoxin onto GlyR ion channels (3,23,24), our electrophysiological data show that ␤⌬7 is not able to efficiently alter ion channel properties.Given the predicted loss of TM2, ␤⌬7 represents a GlyR subunit variant that is no longer able to form a Cl Ϫ channel pore.Rather, ␤⌬7 may act as an accessory subunit of the GlyR complex, potentially contributing to the formation of larger protein complexes and membrane anchoring.
In addition to their classical role as ligand-gated chloride channels of the postsynapse, electrophysiological studies have demonstrated functional expression of GlyR chloride channels (50) and GABA A receptors (51) in astrocytes and other nonsynaptic cells.Due to a lack of synaptic interactions of these cells, ligand-gated ion channels have been proposed to be involved in other functions, such as regulation of cell volume (52), pH (53), and cell proliferation (54).Given their role as interaction partners of gephyrin, GlyR ␤ variants may also contribute to the cellular membrane architecture, independent of the ion channel function (11,55,56).
In addition to the physiological splice variants of the GlyR ␤ subunit gene, pathological alterations in GlyR ␤ pre-mRNA splicing underlie the human neurological disorder hyperekplexia and the spastic mutant of the mouse.In a case of hyperekplexia associated with a compound heterozygotic mutation of the GLRB gene, a splice mutation was found to cause a loss of exon 5 from ␤ subunit transcripts (16).In spastic mice, insertion of a LINE-1 transposable element into intron 5 of the Glrb results in a dramatic reduction in full-length ␤ transcripts (18,57), producing a loss of GlyRs from spinal cord and brainstem.These observations underline the crucial role of the GlyR ␤ subunit for a correct assembly and membrane insertion of GlyRs in vivo.Taken together, it has become apparent that alternative splicing contributes to the subtype heterogeneity of neurotransmitter receptors in the central nervous system.
3 mM MgCl 2 , 1.25 mM K 2 HPO 4 , 26.0 mM NaHCO 3 , 10.0 mM D-glucose, pH 7.4 (a total K ϩ concentration of 5 mM).The solution was gassed with a mixture of 95% O 2 and 5% CO 2 .The internal pipette solution was composed of: 130.0 mM KCl, 0.5 mM CaCl 2 , 3.0 mM MgCl 2 , 5.0 mM EGTA, 10.0 mM HEPES, pH 7.2.Whole cell recordings from transfected HEK293 cells were performed by application of ligand (100 M glycine) or coapplication of ligand together with picrotoxin (100 M) using a U-tube that bathed the suspended cell in a laminar flow of solution, giving a time resolution for equilibration of 10 -30 ms.The external buffer consisted of 137 mM NaCl, 5.4 mM KCl, 1.8 mM CaCl 2 , 1.0 mM MgCl 2 , 5.0 mM HEPES, pH adjusted to 7.2 with NaOH; the internal buffer was 120 mM CsCl, 20 mM N(Et) 4 Cl, 1.0 mM CaCl 2 , 2.0 mM MgCl 2 , 11 mM EGTA, 10 mM HEPES, pH adjusted to 7.2 with CsOH.Current responses were measured at a holding potential of Ϫ60 mV.All experiments were carried out at room temperature.Recording pipettes were fabricated from borosilicate capillaries.The open resistance of these patch pipettes was 5-6 megohms.

FIGURE 1 .
FIGURE 1. Analysis of GlyR␤ subunit expression in glial cells of the rat spinal cord by single cell RT-PCR.A series of spinal cord glial cells were analyzed by patch clamp recording for their expression of glycine-evoked currents and subsequently investigated for the identification of GlyR ␤ subunit encoding mRNAs.A, a series of de-and hyperpolarizing voltage steps (Ϫ160 to 20 mV) from a holding potential of Ϫ70 mV was applied and membrane currents were recorded.The delayed rectifying outward current activated at the depolarizing potential identifies cell number 1 as a precursor cell (P).The symmetrical pattern of in-and outward currents of cell number 2 reflects the current pattern of a mature glial cell (G).B, glycine (1 mM) applied while clamping the membrane potential at Ϫ70 mV evoked a whole cell inward current.C, primer selection for amplification of GlyR ␤ transcripts.For the nested PCR approach two different primer pairs were selected from exons 2, 4, and 8. Using the primer pair bS2 and bAS2 a cDNA fragment of 509 bp for the full-length GlyR ␤ is obtained, whereas cDNAs encoding GlyR␤⌬7 give rise to a truncated fragment of 356 bp lacking the sequence region of exon 7. D, amplified cDNA fragments obtained from single cell RT-PCR of the precursor cell (P) and two mature glial cells (G2 and G3) revealed a length of 356 bp employing GlyR ␤ subunit-specific primers.Lane 4 reflects the negative control of cytoplasmatic RNA obtained from glial cells displaying no glycine-evoked inward current.

FIGURE 2 .
FIGURE 2. Overall topology of GlyR ␤ variants.A, bar diagram representing both ␤ and ␤⌬7 subunits used for in vitro studies.Following the signal peptide (SP), ␤ subunits were N-terminal tagged with a myc epitope (9E10), black boxes represent transmembrane domains 1, 2, 3, and 4. Black line in ␤⌬7 shows localization of exon 7 deletion (amino acid positions 252-301).Gray box marks the binding motif for the anchor protein gephyrin of the GlyR complex at the postsynaptic membrane.B, topology models of the full-length ␤ and ␤⌬7 subunits.Left, the ␤ subunit is a member of the Cys-loop family of receptors with four transmembrane domains.The N terminus is highly homologous to the acetylcholine-binding protein, with a small ␣-helical domain followed by 10 ␤-sheets (arrows).C, proposed model of the ␤⌬7 splice variant.Due to exon deletion, the ion channel domain TM2 is lost and amino acids encoded by exon 8 replace the 10th ␤-sheet.Middle, topology prediction of ␤ and ␤⌬7 based on hydrophobicity analysis.Right, secondary structure prediction for full-length ␤ (Phe 209 - Leu 322 ) and ␤⌬7 (Phe 209 -Leu 272 ) proteins.The amino acid positions of the former TM2 domain of ␤ switch from an ␣-helical structure into a ␤-strand in ␤⌬7.

FIGURE 3 .
FIGURE 3. Expression of ␤ and ␤⌬7 transcripts in various tissues of mice.A, cortex (cx), spinal cord (sc), heart (he), and liver (li) were used for ␤ and ␤⌬7 transcript amplification.␤-Actin (lower panel) was used as housekeeping gene, and reveals that the same amount of RNA was used for each probe.B, transcript level of ␤⌬7 in cultured astrocytes compared with spinal cord neurons.␤ and ␤⌬7 were amplified from plasmids (expected size of 606 and 453 bp) as control for the correct transcript size.Note the high amount of ␤⌬7 transcript observed in cultured glial cells at DIV 14 (cultured at P0).

FIGURE 4 .
FIGURE 4. In vivo protein expression of ␤⌬7 in different tissues.A, membrane preparations from spinal cord (sc), cortex (cx), and liver (li) of adult mice.Positive controls: sc stained with monoclonal anti-gephyrin antibody for recognizing the 93-kDa anchor protein gephyrin (first lane) and with mAb4a to recognize the GlyR ␣1 subunit (second lane).Spinal cord, cortex, and liver were stained using the polyclonal anti-GlyR-␤-H170 antibody, which recognizes an epitope in the intracellular TM3-4 loop of the GlyR ␤ subunit.Cortex preparation was digested with Endo H.The black arrow points to the 58-kDa band of the ␤ subunit; gray arrow to the 53-kDa signal of ␤⌬7.The positive control for Endo H digest is shown in Fig.6B.B, antibody K12 specific for the splice variant ␤⌬7 recognizes an epitope bridging the exon 6/8 border.Membrane preparation from ␤and ␤⌬7-transfected HEK cells, both constructs were stained with anti-myc antibody or K12 (1:100).K12 was tested on liver (li) and mock-transfected HEK cells compared to preadsorbed K12 antiserum.