Single Expressed Glycine Receptor Domains Reconstitute Functional Ion Channels without Subunit-specific Desensitization Behavior*

Background: Functional GlyRα1 receptors can be reconstituted from nonfunctional subunit domains. Results: GlyRα3 and GABAARρ1 rescues are less efficient with the alternative splicing cassette in the TM3–4 loop of GlyRα3 responsible for desensitization and rescue efficiency. Conclusion: Desensitization of GlyRα3 requires nondisrupted intracellular domains. Significance: Independent domain reconstitution does not always recapitulate the full functional properties of receptors. Cys loop receptors are pentameric arrangements of independent subunits that assemble into functional ion channels. Each subunit shows a domain architecture. Functional ion channels can be reconstituted even from independent, nonfunctional subunit domains, as shown previously for GlyRα1 receptors. Here, we demonstrate that this reconstitution is not restricted to α1 but can be transferred to other members of the Cys loop receptor family. A nonfunctional GlyR subunit, truncated at the intracellular TM3–4 loop by a premature stop codon, can be complemented by co-expression of the missing tail portion of the receptor. Compared with α1 subunits, rescue by domain complementation was less efficient when GlyRα3 or the GABAA/C subunit ρ1 was used. If truncation disrupted an alternative splicing cassette within the intracellular TM3–4 loop of α3 subunits, which also regulates receptor desensitization, functional rescue was not possible. When α3 receptors were restored by complementation using domains with and without the spliced insert, no difference in desensitization was found. In contrast, desensitization properties could even be transferred between α1/α3 receptor chimeras harboring or lacking the α3 splice cassette proving that functional rescue depends on the integrity of the alternative splicing cassette in α3. Thus, an intact α3 splicing cassette in the TM3–4 loop environment is indispensable for functional rescue, and the quality of receptor restoration can be assessed from desensitization properties.

The glycine receptor (GlyR) 6 is a ligand-gated chloride channel that mediates fast neuronal inhibition predominantly in adult mammalian brain stem and spinal cord. It is a member of the Cys loop receptor (CLR) superfamily also including the nicotinic acetylcholine receptor, the ␥-aminobutyric acid receptors type A and C, as well as the 5-hydroxytryptamine type 3 receptor. These receptors share a conserved three-dimensional structure with a large extracellular ligand-binding domain followed by four transmembrane domains (TM1-4), connected via intra-and extracellular loops, and a short extracellular C terminus. A functional CLR is composed of five homologous subunits arranged around a central ion-conducting pore (1). The adult GlyR consists of two ␣ and three ␤ subunits (2). The extracellular ligand-binding domain of the glycine receptor harbors two disulfide bonds, one of which is eponymous for the receptor family and a second disulfide bond located in loop C is described for the glycine receptor and the glutamate-gated chloride channel (Glu-Cl) from Caenorhabditis elegans (3,4). The large intracellular loop between TM3 and TM4 (TM3-4 loop, also referred to as ICD) is of highest diversity among CLRs.
Alternative splice sites located within the ICD of GlyR␣1 and GlyR␣3 contribute to GlyR variability. In contrast to the GlyR␣1, splice variants GlyR␣3K (spliced form) and GlyR␣3L (unspliced form) differ in desensitization properties. Homomeric GlyR␣3L ion channels are essentially nondesensitizing, whereas GlyR␣3K desensitizes fast (5). The alternative splicing cassette in ␣3 is composed of 15 residues and carries possible phosphorylation sites. A series of mutations within this insert has shown that amino acids harboring hydroxyl groups (Thr-358/Tyr-367/Ser-370) are important mediators in the desensitization process. In addition, the insert present in ␣3L seems to stabilize the overall spatial structure of the domain thereby regulating receptor gating (6,7). Further studies on the ␣1 TM3-4 loop demonstrated the importance of this domain for forward trafficking, nuclear translocation, and post-translational modifications such as interaction with G␤␥ proteins (8 -10). Common to these various pathways are motifs carrying basic residues at the N-and C-terminal end of the intracellular TM3-4 loop. The importance of the basic stretch 318 RRKRR at the N-terminal end of the TM3-4 loop for trafficking and function of the GlyR␣1 was demonstrated in a study of the mouse mutant oscillator, where a microdeletion generates a premature STOP codon within the alternative splicing cassette of the GlyR␣1 subunit (11,12). This mutation leads to a neuromotor phenotype and death 3 weeks after birth due to the absence of functional adult ␣1 subunits. A recent in vitro study on modular GlyR domain architecture revealed that the function of the truncated oscillator GlyR␣1 protein as well as an equivalent wild type truncated ␣1 can be rescued by co-expression of an independent C-terminal tail construct representing the lacking GlyR␣1 domain. Therefore, functional GlyR␣1 receptors can be rebuilt from independently co-expressed domains (13). Protein truncations have also been observed for GlyRs and GABA A/C receptors being associated with either the human neuromotor disorder hyperkeplexia or a special form of epilepsy GEFS ϩ (14 -16).
Here, we wanted to investigate whether assembly of a functional ion channel from independent domains illustrates a general principle for CLR members not restricted to GlyR␣1. GlyR␣3 and GABA A/C 1, a closely related inhibitory anion channel, were used for intrafamiliar domain complementation among various GlyR subunits (␣1 and ␣3) and interfamiliar rescue between GlyR ␣1 and GABA A/C 1. Furthermore, we show that desensitization is affected by domain complementation and is indeed a measure for the quality of receptor reconstitution.
Our data show that co-expression of receptor domains derived from GlyR␣1 and ␣3 displayed almost mutual compatibility. In contrast, GABA C 1 domains were incompatible with GlyR domains. Functional domain complementation revealed that the TM3-4 loops are major determinants of rescue efficiency. Differences in the desensitization between GlyR␣3 Kand L-splice variants determined by the 15-residue spliced insert within the TM3-4 loop are not maintained by domain co-expressions. Thus, the spliced insert may not be disrupted for restoration of wild type-like receptor properties by subunit complementation.

EXPERIMENTAL PROCEDURES
Sequence Alignment, Homology Modeling, and Building of Hybrid Complexes-Homology models of the transmembrane domain (TM) without loops of ␣1 GlyR, ␣3 GlyR, and 1 GABA A/C were generated by using the crystal structure of the glutamate-gated chloride channel (GluCl␣) from C. elegans at 3.35 Å resolution (Protein Data Bank entry 3RIF (3)) as a template. The sequences of mouse ␣1 GlyR (gi͉118130520_58-1407), mouse ␣3 GlyR (gi͉120300935_32-1474), and rat 1 GABA A/C (gi͉8393398_130-1554) were aligned according to the ClustalW algorithm using the default settings shown in Ref. 17. Additionally, the generated multiple sequence alignment was manually adjusted to coincide with the alignment reported in previous work (18,19). The visualization of the alignment was performed with GenDoc (20). Molecular modeling was performed by standard procedures using MODELLER6.2 (21). All five subunits of the pentamer were modeled simultaneously. The obtained models were improved by 200 steps of conjugated gradient energy minimization using the Powell algorithm in Sybyl7.3 (22). The quality of the models was verified by WHAT_CHECK (23) and DaliLite (24). The visualization of the structures was performed with DS Viewer Pro6.0 (25). Hybrid complexes of ␣1, ␣3, and 1 were generated by replacing side chains of nonconserved TM4 amino acids in the original models using the Sybyl7.3 program package. Noncovalent interactions were improved by 200 steps of conjugated gradient energy minimization using the Powell algorithm.
Molecular Biology and Cloning-Truncated GlyR␣3 and GABA A/C 1 were generated by introduction of an early STOP codon at the corresponding amino acid position to the GlyR␣1 oscillator truncation localized in the TM3-4 loop. The GlyR␣3 was truncated at position L330X and GABA A/C 1 at position Q370X (referring in both cases to numbering of the mature protein). Corresponding tail constructs were generated with no overlapping sequence and represent the lacking portion of both truncated GlyR␣3 and 1 with a large portion of the TM3-4 loop, TM4, and the C terminus. Site-directed mutagenesis was used to exchange the TM4 domains between ␣1 and ␣3 GlyRs. Large overlapping primers representing the specific ␣1 or ␣3 TM4 sequence were used in an overlap extension PCR to amplify extended TM4 domains. These amplimers were cut at subunit-specific restriction sites and ligated afterward into the appropriate tail construct. Chimera between ␣1 and ␣3 were generated using the sequence identity of TM3 and therefore the existence of shared restriction sites. All clones were verified by sequencing and encoded on a pRK5 plasmid under the control of a CMV promoter that allows eukaryotic cell expression.
Membrane Preparation and Biotinylation-Following 24 -48 h post-transfection of various cDNA constructs into HEK293 cells using the calcium precipitation method, cells were harvested for cell lysates, membrane preparations, or used directly for cell surface protein detection with a biotinylation assay. For membrane preparation, the cells were collected in PBS, pH 7.4, centrifuged, and transferred into a potassium buffer containing 10 mM K x H x PO 4 , pH 7.4, 250 mM EDTA, 250 mM EGTA, and protease inhibitor mixture tablets (Roche Applied Science). Cells were homogenized with a glass homogenizer followed by very short sonification and centrifugation at 25,000 ϫ g for 20 min. The pellet was again homogenized, and the centrifugation step was repeated. Pellets were resuspended in buffer B (25 mM K x H x PO 4 , pH 7.4, 200 mM KCl, 250 mM EDTA, 250 mM EGTA, and protease inhibitor mixture tablets; Roche Applied Science) and stored at Ϫ80°C. Biotinylation experiments were performed as described by Unterer et al. (26).
Immunocytochemistry-Transfected HEK293 cells were fixed using 4% paraformaldehyde with 4% sucrose. For intracellular detection of receptor proteins, cells were permeabilized with 0.1% Triton X-100, blocked with goat serum, and stained with the GlyR pan-␣ antibody Mab4a (Synaptic Systems, Göttingen, Germany). All tail constructs were tagged with a Myc epitope and stained using either the c-Myc monoclonal (9E10) or c-Myc polyclonal (C19) antibody (Santa Cruz Biotechnology, Dallas, TX). For detection of the 1 subunit, a polyclonal antibody (kindly provided by R. Enz (27)) was used. This antibody can only be used for immunocytochemistry; however, it does not stain the 1 protein in a Western blot. GlyR␣1 or -␣3 variants were detected with the pan-␣ antibody Mab4a in immunostainings following Western blotting. pDsRed-ER and pDsRed-PM vectors were co-transfected for sub-compartmental localization of ␣3 and 1 variants. Both vectors express fusion proteins of the red fluorescent protein DsRed and a domain of the ER marker calreticulin (pDsRed-ER) or GAP-43, a membrane marker protein.
Electrophysiological Recordings-Whole-cell currents were recorded using a HEKA EPC9 amplifier (HEKA Electronics, Lambrecht, Germany) controlled by Pulse software (HEKA Electronics) on a personal computer. Recording pipettes were pulled from borosilicate glass (World Precision Instruments, Berlin, Germany) using a Sutter P-97 horizontal puller. Ligand application using a U-tube gave a time resolution of 20 -30 ms. The external buffer consisted of NaCl 137 mM, KCl 5.4 mM, CaCl 2 1.8 mM, MgCl 2 1.0 mM, Hepes 5.0 mM, pH adjusted to 7.2 with NaOH; the internal buffer was CsCl 120 mM, N(Et) 4 Cl 20 mM, CaCl 2 1.0 mM, MgCl 2 2.0 mM, EGTA 11 mM, Hepes 10 mM, pH adjusted to 7.2 with CsOH. Current responses were measured at a room temperature of 21-23°C, the holding potential was Ϫ60 mV. For each construct, the mean maximum current (I max ) at a saturating glycine concentration (3 mM) was calculated from all cells that were used for analysis. In recordings with the presence of 1 subunit domains, the ligand GABA was applied at a concentration of 500 M.
Data Analysis-For desensitization analysis, whole-cell current traces were transferred to Microcal Origin (Microcal Software, Inc.), and the decaying current phase was analyzed using a single exponential function plus a constant as shown in Equation 1, I obs ϭ I 1 * e ͑Ϫt⁄ 1 ͒ ϩ I const.
(Eq. 1) where I obs is the observed total current amplitude; I 1 is the fraction of current desensitizing with time constant 1 ; and I const is the amplitude of the nondesensitizing current fraction. For all constructs, a single exponential decay plus a constant term were sufficient to describe desensitization behavior. Functional constants of the co-expressed subunits were compared using one-way analysis of variance (Microsoft Origin) followed by Dunnett's post hoc t test. A probability of error of p Ͻ 0.05 was considered significant.

Generation of Individual Receptor Domains and Chimeric
Constructs-Recent experiments have shown that in vitro functional reconstitution of a truncated GlyR␣1 (␣1trc) receptor, as present in the mouse mutant oscillator, can be achieved by coexpression of the missing C-terminal GlyR (␣1_tail) portion (13). It was demonstrated that two parts of the receptor behave as independent folding domains and are capable of assembling into functional pentameric receptors (Fig. 1A). To test for intersubunit compatibility of truncated N-terminal domains with C-terminal domains derived from other GlyR subunits or even from other members of the CLR family, we cloned the respective truncated and tail constructs with no overlapping sequences for HEK293 cell expression of the following: 1) GlyR␣1 (␣1trc, ␣1_tail); 2) GlyR␣3 (␣3trc, ␣3_tail); and 3) the 1 (1trc, 1_tail) subunit, a member of the GABA A/C receptor family. The GABA A/C 1 subunit as a non-glycine but also an inhibitory Cl Ϫ -permeable receptor isoform of the CLR family was selected, as this subunit was able to form functional homomers in vitro. A variety of different intra- These chimeric tail constructs were generated to analyze the importance of the TM4 in domain complementation, becauseTM4 has been recently shown to play a crucial role for receptor assembly (Fig. 1B) (28). For detection as well as for the analysis of correct membrane integration, a Myc tag was added to the N terminus of the tail constructs.

Subcellular Localization and Expression Levels of Individual
Receptor Domains-Subcellular localization of the truncated N-terminal receptor domains of ␣3trc and 1trc was investigated using single expression as well as co-expression with the appropriate C-terminal portion. Single domains of ␣3 or 1 and the corresponding wild types were co-expressed with pDsRed-ER (encoding a fusion protein of DsRed and calreticulin as an ER marker) or pDsRed-PM (fusion protein of DsRed and GAP-43 for membrane expression) to distinguish between intracellular and surface receptor proteins.
The monoclonal antibody Mab4a was used for detection of the GlyR truncated domain ␣3trc, and the polyclonal anti-1 antibody was used to stain the GABA A/C 1 truncated N-terminal domain (1trc). Detection of the tail constructs was performed using a monoclonal or a polyclonal c-Myc antibody depending on the origin of the antibody against trc variants. Because of the intracellular location of the Myc epitope attached to the C-terminal tail variants, all stainings were carried out after permeabilization of transfected HEK293 cells.
For both the GlyR␣3L as well as the GABA A/C 1 wild type (WT) receptors, staining throughout the cell body could be observed with expected enhanced fluorescence at the cell membrane (Fig. 2, A and B). The truncated N-terminal domains ␣3trc and 1trc, in contrast, showed a high intracellular fluorescence with no detectable accumulation at the plasma membrane, indicating a disturbed subcellular trafficking of truncated variants (Fig. 2, A and B, middle panels). The same pattern of distribution was observed in previous studies on ␣1trc (13). The exclusively expressed tail constructs (␣3_tail and 1_tail) displayed a strong cytoplasmic staining, presumably due to retention of most of the protein within the endoplasmic reticulum (Fig. 2, A and B, lower right panels). Thus, no robust localization at the plasma membrane could be observed for any of the single expressed receptor domains of GlyR␣3 and 1.
Upon co-expression of the truncated ␣3 and 1 together with ␣3_tail or 1_tail domains, however, co-localization as well as translocation to the plasma membrane of the complementary fragments were observed (Fig. 2, C and D, merged picture with insets demonstrating close proximity to the membrane marker GAP-43). These findings could be corroborated using Western blotting of whole-cell lysates (Fig. 3A) and specific cell surface localization using biotinylation of protein as a tool for separation from intracellular polypeptides (Fig. 3, B-D). In contrast to ␣3 variants, 1_tail expression was highly decreased in wholecell as well as surface expressions (Fig. 3, C and D).

Differences in Rescue Efficiency between Intra-and Intersubunit Rescue Combinations-The
GlyR␣1 is composed of independent folding domains able to restore functionality from nonfunctional subunit domains (13). Indeed, co-expression of ␣1trc with ␣1_tail resulted in glycine-gated currents of up to 1.4 Ϯ 0.2 nA, which corresponds to about 50% rescue efficiency compared with Cl Ϫ currents recorded from ␣1 WT expression in vitro (3.0 Ϯ 0.6 nA, Fig. 4, A and B, and Table 1). Co-expressions of ␣3 and 1 domains restored functionality of glycinegated ␣3 and GABA-gated 1 channels with efficiencies of 16% for ␣3 and 7.5% for 1 compared with the appropriate WT currents (Fig. 4, A and B, and Table 1). For test of intersubunit compatibility of domains for functional ion channel rescue, ␣1trc was co-expressed with ␣3_tail or 1_tail and ␣3trc together with ␣1_tail (Fig. 4A). Interestingly, the ␣3trc was rescued by the ␣1_tail (0.26 Ϯ 0.1 nA with 8% of ␣3 WT), but restoration of ␣1trc did not occur in a co-expression with the ␣3_tail nor with the 1_tail (Fig. 4C). Thus, these data show that  (␣3trc, 1trc); boldface letters were used to mark the construct used for analysis. Data were pooled from three independent experiments. pan-Cadherin was used as the housekeeping protein expression control (see also C and D). C and D, representative Western blots from biotinylation experiments. C, whole-cell proteins were stained with either the pan-␣ GlyR antibody Mab4a (␣3 WT and trc) or the c-Myc antibody (for ␣3 and 1 tail constructs). pan-Cadherin served as expression control. D, surface proteins labeled. Again, pan-cadherin was used as housekeeping protein (upper panel, 125 kDa). Although variants of 1 were transported to some extent to the cell surface, 1 tail domains showed a decreased protein expression compared with GlyR␣3. A staining of 1 WT or 1trc was not possible due to lack of a specific antibody able to detect 1 protein upon Western blotting.

Domain Architecture of GlyR␣1 and GlyR␣3
OCTOBER 17, 2014 • VOLUME 289 • NUMBER 42 an interfamiliar functional complementation of independent folding domains seems to be impossible.
To verify whether differences in the TM3-4 loop or TM4 are the underlying cause for the lack of functional rescue of ␣1trc with the ␣3_tail, we used molecular modeling of the TM domains to determine surface contacts of TM4 to TM1-3 based on the structural information of the GluCl (Fig. 6A). TM1 and TM3 exhibit identical amino acids at the interface in ␣1 and ␣3 that is recognized by TM4 (Fig. 6B, dotted line). The side chains of residues ␣1Ile-270 localized at the interface in TM1 and of ␣1Gly-284 in TM2 twisted outward from the contact interface of ␣3TM4 to ␣1TM1-3 (Fig. 6B). The in silico analysis suggested that this interface, which is composed of identical amino acids in ␣3 and ␣1, can bind both the TM4 of ␣1 and ␣3  A, representative traces of intrafamiliar functional rescue experiments of ␣1trc ϩ ␣1_tail, ␣3trc ϩ ␣3_tail, 1trc ϩ 1_tail, and interfamiliar domain co-expressions of ␣1 variants with ␣3 and vice versa, also ␣1 with 1 (ratio 1:5 of Xtrc:X_tail with X either ␣1, ␣3, or 1). Glycine (500 M or 3 mM) was applied for 2 s or 10 s as indicated. When 1trc was expressed, GABA-induced (500 M and 3 mM) GABAergic currents were detected. B, relative maximal current amplitudes (I max ) from intrafamiliar rescue experiments are shown. The mean I max values of the wild type subunits ␣1, ␣3L, and 1 was set to 100%. Note, the reduced rescue efficiency for ␣3 and 1 complementation from independent domains is compared with ␣1. C, relative I max values from interfamiliar rescue experiments (trc and tail domain from different GlyR subunits or GlyR␣1 together with 1) compared with intrafamiliar rescue (trc and tail domains originate always from the same subunit). The mean current amplitudes of the most efficient rescue of ␣1 were set to 100%. Interestingly, the ␣3trc is rescued by an ␣1_tail but ␣1trc together with ␣3_tail does result in nonfunctionality.

Domain Architecture of GlyR␣1 and GlyR␣3
OCTOBER 17, 2014 • VOLUME 289 • NUMBER 42 and is therefore most likely not responsible for the nonfunctionality between ␣1trc and ␣3_tail.
Thus, the presence of a TM3-4 loop of the same GlyR subunit as the co-expressed truncated N-terminal receptor portion seems to improve GlyR domain-domain communication. Although our models demonstrated no differences at the contact sites of TM4 between ␣1 and ␣3 toward other TMs of the same subunit, TM4 of ␣3 led to an increase in I max values upon co-expression with ␣3trc.
Chimeric Complementation of GlyR␣1 and GlyR␣3-The ␣3_tail does not coassemble with ␣1trc. Chimeric approaches between various types of CLRs, however, have demonstrated that exchanges of the TM3-4 loop between different members of the CLR family do indeed result in functional receptors. Even more, ion channel properties determined by the TM3-4 loop sequence were restored and transferable to other subunits (31,32). We generated chimeric constructs of ␣1 and ␣3 domains that have been used for domain complementation experiments before to analyze the function of both domains in a continuous polypeptide chain (Fig. 8A). The ICD of ␣1 and ␣3 harbors alternative splicing cassettes that vary in length (8 amino acid residues in ␣1 and 15 amino acid residues in ␣3, Fig. 5) (5, 33). Different from ␣1 splice variants, the long ␣3 receptor variant ␣3L differs to the short variant ␣3K in desensitization (5).
All chimera were expressed at the outer cell surface of transfected nonpermeabilized HEK293 cells (Fig. 8B). Whole-cell recordings in the presence of saturating glycine concentrations (3 mM) revealed functional ion channels for all ␣1-␣3 chimera with I max values not significantly different from ␣1 or ␣3 WTs ( Fig. 8C and Table 2).
Traces of ␣3K, ␣3L, and the ␣1␣3 chimeras were analyzed for desensitization kinetics (Fig. 8D) due to the fact that the subunit switch lies within the subdomain determining differences in desensitization in GlyR␣3. Lack of the spliced insert in ␣3K resulted in desensitization time constants of 1.5 Ϯ 0.2 s similar to previously published data (5,7). The desensitization properties of ␣3K and ␣3L are restored in the ␣1␣3 chimera ( Fig. 8D and Table 2). The ␣1rescue construct harboring nine residues of the 15-amino acid insert present in ␣3L ended up with a large fraction of nondesensitizing currents (␣1rescue 75.2% compared with ␣3L 91.3%). These data show that not all residues of the 15-amino acid splice cassette are required to generate nondesensitizing ion channels.
Desensitization in Domain Complementation Studies-Here, we used ␣3trc and co-expressed this domain with various ␣3_tails to analyze whether desensitization is restored in an ion channel assembled from independent domain co-expressions. Tail constructs with the complete splice cassette (␣3L_tail), a half-splice cassette (␣3_tail), and lack of the insert (␣3K_tail) were used (Fig. 9A). The expression level of the tail constructs together with ␣3trc was verified in a biotinylation assay. Although all GlyR domains are expressed in total protein samples, the surface expression varied considerably (Fig. 9B). The ␣3trc and ␣3_tail expressions in the plasma membrane were markedly reduced compared with the other constructs. This finding might also explain the low rescue efficiency observed in whole-cell recordings for the ␣3trc ϩ ␣3_tail (Fig. 9B, lane 2) co-expression. The expression of the ␣3K_tail and ␣3L_tail was higher, leaving the low cell surface expression of ␣3trc unaffected (Fig. 9B). Functional ion channels were generated by all domain co-expressions of ␣3trc with various ␣3_tails generat-ing I max reductions of 8 -43% of ␣3L WT that might be due to the low expression levels of ␣3trc (Fig. 9, B and C). Again, ␣1trc was inefficiently rescued by either the ␣3K_tail or the ␣3L_tail (Fig. 9C) arguing for poor interaction between the N-terminal domain of GlyR␣1 with the C-terminal domain of GlyR␣3 in a co-expression approach. The fast desensitization of ␣3K was completely abolished by domain co-expressions. All restored ␣3 ion channels showed nondesensitizing current responses with time constants similar to ␣3L (Table  2). Likewise, the fraction of nondesensitizing currents was FIGURE 8. Chimeric glycine receptors keep desensitization properties. A, chimeric receptors used to assess mutual compatibility of domains derived from different subunits. Transition of the subunits is between TM3 and TM3-4 loop. Note that in some cases only the TM3-4 loop is exchanged. B, nonpermeabilized cells were stained with the monoclonal Mab2b antibody to evaluate surface expression. All chimeric receptors trafficked to the cell surface with no major differences compared with the respective ␣1 WT. C, maximal glycine-evoked currents (I max ) recorded in whole-cell recordings from HEK293 cells expressing the chimeric GlyRs. ␣3K and ␣3L represent the two splice variants of ␣3 wild type. In ␣1-␣3L and ␣1-␣3K, the ligation between ␣1 and ␣3 was done at the end of TM3. ␣1rescue corresponds to a construct harboring the first 21 amino acid residues of the TM3-4 loop from ␣1 (corresponding to the ␣1 oscillator variant) followed by the remaining TM3-4 loop residues, TM4, and the C terminus of ␣3. The last two bars at right results from the chimera that contained only the TM3-4 loop either at ␣3K or ␣3L in an ␣1 surrounding. All of the receptor variants responded to saturating glycine concentrations (3 mM) with large inward chloride currents (n Ն 7). No significant differences compared with the wild type could be observed (analysis of variance combined with post hoc t test). D, representative current traces of ␣1␣3 chimera and ␣3K and ␣3L used for calculation of desensitization behavior. The fraction of nondesensitizing currents of the ␣1␣3 chimera was compared with ␣3L (nondesensitizing) and ␣3K (desensitizing) WT (for values see Table 2). Significance was determined using a one-way analysis of variance; *, values of p Յ 0.05 are considered significant; **, p Յ 0.01; ns, not significant.

TABLE 2 Desensitization properties of various co-expressed GlyR␣3 domains
HEK293 cells expressing different GlyR␣3 variants without or with co-expressed tail domains were patched following 48 h post-transfection; n ϭ number of cells recorded; agonist concentration was applied 3 mM glycine; ϩ , n ϭ number of cells responded out of all cells recorded; #, only one cell. p values refer to the appropriate WT values as follows: *, p Ͻ 0.05; ***, p Ͻ 0.001.

Expressed domains
No. of cells I max (pA) ؎ S.E.   Table 2). The constructs are indicated at the bottom of the bars. Differences between ␣3L WT and functional receptors generated from independent domain expressions were tested for significance using one-way analysis of variance. D, fraction of nondesensitizing currents (means Ϯ S.E.) observed for ␣3L WT and ␣3trc co-expressed with tail constructs harboring or lacking the splice cassette important for desensitization of ␣3 (for values see Table 2). Representative current traces of ␣3trc and various ␣3_tail constructs used for calculation of desensitization behavior are shown on the right. Value # was calculated only from one recorded trace; out of 19 cells recorded (␣1trc ϩ ␣3L_tail) only two responded to the agonist glycine, see also Table 2. Significance was determined using an one-way analysis of variance combined with Dunnett's post hoc t test, with values shown as follows: ***, p Յ 0.001; ns ϭ not significant.

Nondesensitizing current fraction (%) ؎ S.E. desens (s) ؎ S.E.
not altered between ␣3L WT and domain complementations ( Fig. 9D and Table 2). Hence, desensitization of GlyR␣3 is not only determined by the presence or absence of the alternative splicing cassette; an interaction of the TM3-4 loop subdomain around the 15-residue cassette with other GlyR domains is required to enable ion channel desensitization, which is most likely disturbed by the domain co-expression approach.

DISCUSSION
Ligand-gated ion channels are composed of independent subunits. Each subunit is divided into distinct domains. A recent study on the GlyR␣1 normally leading to death of homozygous mice at the end of the 3rd postnatal week revealed that co-expression with the lacking portion of ␣1 restored the expression of the truncated GlyR␣ protein in neurons and ion channel function in vitro in transfected cell lines (13). Hence, GlyR␣1 receptors are composed of independent folding domains generating functional ion channels from nonfunctional subunit fragments. Similarly, co-expression of single domains of the muscarinic m3 acetylcholine receptor or the GluN1/GluN2A receptor resulted in functional rescue of single nonfunctional domains (34,35).
In the inhibitory GlyR, the mouse mutant oscillator serves as a model system for human hyperkeplexia (12,36), because similar truncations have been found in human patients that suffer from this neuromotor disorder. The observed phenotype of these patients is most probably due to a disturbed expression of affected ␣1 subunits (14,16). Truncations of proteins have also been detected in other channelopathies, such as cystic fibrosis or a special form of epilepsy (GEFS ϩ ) (37,38).
To test whether the in vitro rescue of the oscillator defect is transferable to other CLRs, we introduced corresponding truncations into the highly homologous subunit GlyR␣3 and another closely related Cl Ϫ channel, the 1 GABA A/C receptor. The oscillator truncation in GlyR␣1 is localized within the large ICD in an alternative splicing cassette (39), giving rise to two splice ␣1 variants in wild type mice that do not differ in the extent of desensitization and other functional properties. Independent of alternative splicing, the large ICD is of highest variability among CLRs and carries important motifs for phosphorylation, interaction with cytoskeletal proteins or other interaction partners (40 -42).
For GlyR␣1, amino acid residues have been identified that are essential for proper assembly and receptor pentamerization (9,28,43). The functional rescue shown for GlyR␣1 was highly dependent in efficiency on the basic stretch 318 RRKRR, which is essential for surface expression. GlyR␣3 contains the same basic subdomain at the N terminus of the ICD. This motif is unaffected by the truncation corresponding to ␣1 oscillator.
The functional rescue of truncated GlyR␣3 and GABA A/C 1 in the TM3-4 loop showed decreased rescue efficiencies of 16 and 7.5% compared with WT activities. An interfamiliar cross of ␣1 with 1 and vice versa never led to the formation of functional ion channels indicating that these domains are incapable of coassembling into heteromeric GlyR-GABA C receptors. The subtle changes of intersubunit interactions appear sufficient to prevent efficient assembly of mixed GlyR/GABA C receptor channels.
Interestingly, our findings of intrafamiliar rescue between GlyR␣1 and GlyR␣3 led to functional reconstitution only in one direction with truncated ␣3 together with an ␣1 complementation domain. Why is the ␣3_tail not able to complement ␣1trc? First, ␣1trc might hinder transport of ␣3_tail to the cell surface. Second, the position of the truncation in the ␣3_tail lies within an alternatively spliced 15-residue motif critical for receptor desensitization (5,6). Third, a different positioning of TM4 in ␣3 disables communication with TM1-3 of ␣1 (40). Consistent with these assumptions, we have detected a low expression of the ␣3_tail upon co-expression with ␣1trc. The ␣1trc construct harbors the membrane-directing motif 318 RRKRR and incorporates well into the plasma membrane alone (13) and in co-expression with ␣3_tails.
The truncations corresponding to the oscillator mutation in ␣1 and ␣3 lie in alternative splicing cassettes. In ␣1, the truncation does not hinder the interaction of front and tail fragments to restore functionality. A recent study showed that upon co-expression of GlyR␣1 domains, an N-terminal truncation of more than 49 residues in the TM3-4 loop of GlyR␣1 resulted in nonfunctional channels (26).
In the ␣3 sequence, the 15-amino acid insert determines differences in desensitization of both ␣3 variants (5). The truncation does result in two ␣3 domains both harboring half of the alternative splicing cassette. This disruption might disable communication with the ␣1trc.
Structural information of the TM3-4 loop that might explain intrasubunit interactions is scarce, except for the short MA-stretch N-terminal to TM4 (44,45). Sequence comparison and modeling indicate that the TMs in ␣1 and ␣3 are almost identical. TM4 carries six different amino acid residues, three of which are nonconserved between ␣1 and ␣3. An exchange of these residues between ␣1 and ␣3 has been proposed to be responsible for disparities in agonist efficacies mediated by diverse interactions with the surrounding lipids. These variations in glycine efficacy of ␣1 to ␣3 are due to different orientations of TM4 to their corresponding TM3 domains (31,40). Aromatic residues in TM4 that are important for pentamerization of GlyR complexes are conserved between ␣1 and ␣3, e.g. in ␣1Phe-425, Phe-429, Phe-432, Phe-435, Tyr-436, Trp-437, and Tyr-440 (28). Thus, a defect in pentamerization is most likely not the main determinant for lack of functional reconstitution from co-expressed GlyR domains.
The contact surface of TM4 toward TM1-3 is identical between ␣1 and ␣3. The two residues, which are different between ␣1 and ␣3, are not directed toward TM4. Using tail variants with exchanged TM4s of ␣1 or ␣3 to generate a domain co-expression with subunit-specific TM4s generated functional ion channels for ␣3trc with both ␣1_tails either containing a TM4 of ␣1 or ␣3. Hence, TM4 is interchangeable without loss of function. Rather, the disruption of the TM3-4 loop sequence within the alternative splicing cassette of the ␣3 results in a defective subdomain important for domain-domain communication.
This alternative splicing cassette with additional 15 amino acids in ␣3 resulting from an unused alternative splice acceptor site has been shown to determine desensitization of the two resulting ␣3 variants (5, 6). Three hydroxylated amino acids Domain Architecture of GlyR␣1 and GlyR␣3 OCTOBER 17, 2014 • VOLUME 289 • NUMBER 42 within the insert of 15 residues in the TM3-4 loop of ␣3 have been intensively studied as crucial determinants of desensitization. Furthermore, the insert seems to stabilize a spatial structure of the domain (6,7). Therefore, the truncation localized within the ␣3 insert might disturb desensitization. In our complementation experiments, no differences in desensitization time constants were observed.
Does this necessarily mean that desensitization behavior in general cannot be transferred between subunits? In contrast to domain co-expression, disruption of the 15-residue insert in chimera of ␣1 and ␣3 gave rise to large fractions of desensitizing currents. The presence of the insert led to nondesensitizing currents. These data demonstrate that the chimera between ␣1 and ␣3 results in functional ion channels similar to previous findings on the exchange of the full-length TM3-4 loop plus TM4 between both GlyR subunits (31). Hence, in a continuous polypeptide chain, desensitization can be transferred from ␣3 to ␣1 constructs via a TM3-4 loop switch between these subunits. In contrast, desensitization is not transferable in domain co-expression experiments, although functionality of the receptor was restored. The disrupted ␣3 insert localized at the N terminus of the tail domain was described to stabilize the secondary structure of the GlyR␣3 TM3-4 loop (6). The N-terminal location of the 15-residue insert might disable the formation of secondary structures and requires neighboring residues that facilitate a structural stabilization of the domain necessary for receptor conformations and re-orientation of TM3 during desensitization.
Thus, receptor desensitization was identified as a useful measure for the quality of ion channel assembly as it allows us to distinguish between continuous polypeptides and receptors assembled from subunit fragments. Overall, these data provide further evidence that desensitization and channel opening are separate elementary processes that are governed by distinct and structurally well defined domains of the receptor protein.