Electrophysiological Signature of Homomeric and Heteromeric Glycine Receptor Channels*

Glycine receptors are chloride-permeable, ligand-gated ion channels and contribute to the inhibition of neuronal firing in the central nervous system or to facilitation of neurotransmitter release if expressed at presynaptic sites. Recent structure-function studies have provided detailed insights into the mechanisms of channel gating, desensitization, and ion permeation. However, most of the work has focused only on comparing a few isoforms, and among studies, different cellular expression systems were used. Here, we performed a series of experiments using recombinantly expressed homomeric and heteromeric glycine receptor channels, including their splice variants, in the same cellular expression system to investigate and compare their electrophysiological properties. Our data show that the current-voltage relationships of homomeric channels formed by the α2 or α3 subunits change upon receptor desensitization from a linear to an inwardly rectifying shape, in contrast to their heteromeric counterparts. The results demonstrate that inward rectification depends on a single amino acid (Ala254) at the inner pore mouth of the channels and is closely linked to chloride permeation. We also show that the current-voltage relationships of glycine-evoked currents in primary hippocampal neurons are inwardly rectifying upon desensitization. Thus, the alanine residue Ala254 determines voltage-dependent rectification upon receptor desensitization and reveals a physio-molecular signature of homomeric glycine receptor channels, which provides unprecedented opportunities for the identification of these channels at the single cell level.

Glycine receptors (GlyRs) 5 are chloride-permeable, ligandgated ion channels (LGICs) that contribute to synaptic communication in the central nervous system (1,2). GlyRs are involved in locomotion and the processing of visual, acoustic, and sensory signals. Aberrant function is associated with neuropathic pain and hyperekplexia (startle disease) as well as the pathophysiology of temporal lobe epilepsy and autism spectrum disorder (3)(4)(5)(6)(7). The family of GlyRs comprises five genes in humans: GLRA1-4, coding for the subunits GlyR ␣1-4, and GLRB, coding for GlyR ␤ (2,8). Each GlyR subunit is composed of four transmembrane domains (TM1-4) and extracellular N and C termini. Amino acids contributing to the glycine binding site are located in the large extracellular N-terminal domain, whereas the intracellular linker between TM3 and TM4 hosts several signaling domains and phosphorylation sites (8). Functional GlyR channels are either homopentamers formed by five identical ␣ subunits or heteropentamers formed by ␣ and ␤ subunits at a 2:3 stoichiometry (9, 10). GLRA4 is a pseudogene (8). Compared with other members of the LGIC superfamily, the subunit diversity of GlyRs is rather moderate, but posttranscriptional modifications generate a broad diversity. Alternative splicing of the large TM3-TM4 loop in ␣1 subunits leads to two GlyR ␣1 variants, ␣1Ins and ␣1⌬Ins, which differ from each other in the presence or absence of an eight-amino acid insert (SPMLNLFQ), respectively (11). Likewise, two ␣3 variants, ␣3K and ␣3L, are generated by alternative splicing of GlyR ␣3 exon 8A (TEAFALEKFYRFSDT) located in the TM3-TM4 loop (12). For GlyR ␣2, alternative splicing of exon 3 generates two different receptor variants that differ by two amino acids (␣2A (IA) and ␣2B (VT)) in the extracellular ligand binding domain (13).
To investigate GlyRs on a splice-specific level, detailed information about functional properties of each channel variant is needed. The electrophysiological properties of GlyR channels have been studied in various overexpression systems, including HEK293 cells and Xenopus laevis oocytes. Most likely due to the different cellular expression systems and cellular intrinsic signaling profiles, the apparent glycine affinity of the different homo-and heteromeric channel variants, for example, varies largely between 10 and 300 M (8, 14 -18). Hence, a study that compares all the GlyR variants in the same cellular context and using the same experimental setup is required to provide unambiguous information about the functional properties of different GlyR isoforms. A hallmark of all LGICs, including GlyRs, is the desensitization of currents in the continuous presence of a ligand, a mechanism that assures and shapes fast synaptic transmission (19). GlyR desensitization is a complex process that involves slow and fast decay components (1). Compared with the closely related GABA type A receptors (GABA A R), GlyRs desensitize relatively slowly (0.5-11 s) (1,12,18). Here again,  time course and decay half time of desensitization underlie large cell-to-cell variability (20) due to many different parameters that may influence the kinetics, such as ligand concentration (16), the phosphorylation state of the intracellular TM3-TM4 loop (16,20), or the membrane potential (1,21,22). GlyR are basically selective for Cl Ϫ but also permeable to other anions and even to cations (23)(24)(25)(26)(27)(28). Due to anomalous mole fraction effects on ion conductance, it has further been suggested that GlyRs are multi-ion channels with at least two ion binding sites in the channel pore (23,24). The GlyR pore is formed by the TM2 helices from five channel-forming GlyR subunits. The narrowest point within the pore is located at the intracellular pore mouth and formed by the 2Ј-Pro and 9Ј-Leu residues within TM2 (29). Interestingly, amino acids from this region are involved not only in selectivity of GlyR to Cl Ϫ , but also in determining its desensitization kinetics (18). A current-voltage (IV) relationship characterizes and identifies an ion channel. Characteristic IV relationships were identified for many homo-and heteromeric ion channels, including members of the transient receptor potential (TRP) channel family (30,31). However, in the case of GlyRs, detailed information about isoform-specific characteristics of IV relationships is not available, especially for ␣2and ␣3-containing GlyRs that are involved in neuropathic pain, temporal lobe epilepsy, and autism spectrum disorder (3)(4)(5)(6). Here, using the same (HEK293T) cellular expression system and the same experimental setup, we reveal essential information about characteristic IV relations of recombinant homomeric GlyR ␣2 and ␣3 channels. The results show that, upon receptor desensitization, the IV relationship switches from a linear behavior to an inwardly rectifying shape, in contrast to homomeric GlyR ␣1 or heteromeric GlyR ␣2/␤ or ␣3/␤ channels, where IV relation-ships remain linear. Furthermore, GlyR ␣2 and ␣3 IV rectification is dependent on a single amino acid at the inner pore mouth and closely linked to Cl Ϫ permeation. This characteristic electrophysiological IV signature of homomeric GlyR ␣2 and ␣3 channels was also found in primary hippocampal neurons with preponderant GlyR ␣2 protein expression. Thus, our study not only provides useful information about the physiomolecular signature of GlyR channels and their identification in native cells but also may help in understanding the complex nature of maladaptive forms of neuronal plasticity in neuropathic pain, autism spectrum disorder, and temporal lobe epilepsy.

Current-Voltage Relationships of Desensitized GlyR ␣3L
Are Inwardly Rectifying-The electrophysiological properties of GlyR ␣3L were investigated in whole-cell patch clamp recordings on HEK293T cells transiently transfected with the cDNA of the receptor subunit. In agreement with previous publications, application of glycine elicited inward currents at Ϫ50 mV in a concentration-dependent manner (Fig. 1A). The time courses of glycine-evoked currents were characteristic in terms of a fast increase that was followed by a slow decline toward a steady state level, which is referred to as receptor desensitization. Note that although considered as inward current by convention, the Cl Ϫ flux through GlyR ␣3L at Ϫ50 mV is actually directed outwardly. When glycine was removed from the bath solution, currents returned to baseline levels. The specificity of glycine-induced currents through GlyR ␣3L was further confirmed by application of the competitive antagonist strychnine (10 M), which completely and reversibly blocked GlyR ␣3L  (1)(2)(3)(4) are shown in D. D, IV relationships of homomeric GlyR ␣3L channels during current desensitization in response to 1000 M glycine. Note that the IV relationships at peak are linear and become inwardly rectifying upon desensitization. E, averaged IV relationships of homomeric GlyR ␣3L channels upon desensitization and in response to 1000 M glycine (n ϭ 12). All IV relationships were normalized to the currents at Ϫ150 mV before merging. F, averaged IV relationships of homomeric GlyR ␣3L channels upon desensitization and in response to 100 M glycine (n ϭ 10). All IV relationships were normalized to the currents at Ϫ150 mV before merging and to the peak currents of responses at 1000 M glycine after merging. G, summary of the potentials at which the maximal IV outward currents were determined (V(I max )), for currents evoked with 30, 100, and 1000 M glycine. H, summary of the rectification index (ϪI Ϫ150 mV /I ϩ150 mV ) for currents through GlyR ␣3L homomers evoked with 30, 100, and 1000 M glycine. Error bars, S.E. responses to 1000 M glycine (Fig. 1B, bottom). Non-transfected HEK293T cells did not respond to glycine (Fig. 1B, top).
IV relationships of GlyR ␣3L were determined at different time points of receptor activation by applying voltage ramps ranging from Ϫ150 to ϩ150 mV (600 ms) every 5 s. As expected for a Cl Ϫ channel, each IV relationship of GlyR ␣3L reversed close to the Nernst potential for Cl Ϫ (Ϫ2.3 mV). Instantly after glycine administration, IV relationships of GlyR ␣3L were linear ( Fig. 1, C and D) but, interestingly, became inwardly rectifying upon desensitization. To quantify this rectification, we analyzed IV relationships obtained 25-30 s after glycine application, a time point at which homomeric GlyR ␣3L channels were in a steady state because the time constant of desensitization was 5.3 Ϯ 1.1 s (n ϭ 13; monoexponential fit of time courses at Ϫ50 mV; see Table 1). The following parameters were extracted from these IVs: ϪI ϩ150 mV /I Ϫ150 mV , which is a measure of the extent of IV rectification, and V(I max ), which gives the membrane potential at which the outward currents start to decline. Thus, a channel displaying a perfectly linear IV relationship (reversing at 0 mV), which exerts no rectification and no decline in conductance at positive potentials, had parameters of ϪI ϩ150 mV /I Ϫ150 mV ϭ 1 and V(I max ) ϭ 150 mV. For desensitized GlyR ␣3L, ϪI ϩ150 mV /I Ϫ150 mV was 0.17 Ϯ 0.04 (n ϭ 12), and V(I max ) was 80.9 Ϯ 8.2 mV (n ϭ 12), indicating that the outward currents at ϩ150 mV were nearly 7 times smaller than the inward currents at Ϫ150 mV and that this voltage-dependent decrease in conductance started at around ϩ75 mV ( Fig. 1D and Table 2). Interestingly, after submaximal activation of GlyR ␣3L with only 30 or 100 M glycine, the IV inward rectification and the decline of conductance at positive potentials disappeared, and the IV relationships were rather linear ( Fig. 1, E-H).
Determinants of GlyR ␣3L Rectification-We investigated whether an open channel block by intracellular Mg 2ϩ or Ca 2ϩ could be responsible for the reduced conductance of GlyR ␣3L at positive membrane potentials. A similar mechanism has been elucidated for other ion channels e.g. inward rectifier potassium channels (K ir ), where inward rectification depends on a highly voltage-dependent block by intracellular Mg 2ϩ and organic polyamines (32)(33)(34). However, in the case of GlyR ␣3L, there was no change in the shape of the IV relationships when using Ca 2ϩ -and Mg 2ϩ -free pipette solutions containing 10 mM HEDTA (Fig. 2, A and C). Consequently, rectification (ϪI ϩ150 mV /I Ϫ150 mV ϭ 0.21 Ϯ 0.10, n ϭ 4; p ϭ 0.5899) and V(I max ) (79.9 Ϯ 23.5 mV, n ϭ 4; p ϭ 0.9591) were similar to those in the experiments with standard pipette solution, indicating that an open channel block by intracellular divalent cations is unlikely to be responsible for the inward rectification.
To study the mechanisms of GlyR ␣3L rectification further, we considered the possibility that Cl Ϫ permeation through GlyR ␣3L is limited at positive potentials. To address this option, we performed recordings on GlyR ␣3L using low intracellular Cl Ϫ concentrations and applied 1000 M glycine. As shown in Fig. 2, B and C, inward rectification was still present under these conditions; however, V(I max ) (28.1 Ϯ 1.1 mV, n ϭ 5) was strongly shifted toward more negative potentials. Intriguingly, when compared with high Cl Ϫ pipette solutions, the shift of V(I max ) was similar to the shift of the reversal potential (Ϫ42.0 Ϯ 2.1 mV, n ϭ 5; Fig. 2C), indicating that the decrease in GlyR ␣3L conductance at positive membrane potentials is related to diminished Cl Ϫ permeation at these membrane potentials.
IV Rectification Identifies Homomeric GlyR␣2 and GlyR␣3 Channels-We next studied the characteristics of different RNA splice variants of homomeric GlyR ␣1, GlyR ␣2, and GlyR ␣3 channels and compared their major properties with those of GlyR ␣3L. As shown in Fig. 3, each of the investigated isoforms responded to the application of glycine. All data on concentration-response curves and desensitization kinetics can be found in Table 1, and the parameters extracted from IV relationships are summarized in Table 2. Like GlyR ␣3L, the IV relationships of homomeric GlyR ␣2and ␣3-containing channels were inwardly rectifying upon desensitization, whereas GlyR ␣1 channels displayed linear IV relationships. We also investigated GlyR ␣2/␤ and GlyR ␣3/␤ heteromeric channels. The ␤ subunit was previously shown to influence not only the subcellular expression of GlyRs but also their biophysical properties. In our experiments, when comparing homomeric and heteromeric GlyR, the ␤ subunit-containing channels generally tended to have lower affinities (EC 50 ) and smaller cooperativities (n H ) for Electrophysiological properties of GlyR channels at ؊50 mV I 1000 gly, peak , average amplitudes of peak currents through the indicated GlyR channels in whole-cell patch clamp recordings at Ϫ50 mV evoked by 1000 M glycine. EC 50, peak and n H, peak , EC 50 values and Hill coefficients of peak currents through the indicated GlyR channels in whole-cell patch clamp recordings at Ϫ50 mV. 1000 gly , averaged time constants of decay of currents through the indicated GlyR channels evoked by 1000 M glycine. Values were determined by a monoexponential fit of the time courses at Ϫ50 mV. I 1000 gly, desens. , Average amplitudes of peak currents through the indicated GlyR channels in whole-cell patch clamp recordings at Ϫ50 mV evoked by 1000 M glycine. EC 50, desens. and n H, desens. , EC 50 values and Hill coefficients of desensitized currents through the indicated GlyR channels in whole-cell patch clamp recordings at Ϫ50 mV. Current amplitudes were extracted from the time courses at Ϫ50 mV 25-35 s after glycine application. Numbers in parentheses indicate the number of experiments (n).
Functional Determination of Somatic GlyR Subunit Composition in Hippocampal Neurons-We finally asked whether the particular biophysical characteristics of GlyR ␣1 versus ␣2 and ␣3 can be used as an electrophysiological signature or fingerprint of the neuronal endogenous GlyR subunit composition. It was recently shown that the GlyR ␤ subunit is expressed as mRNA, not as protein, in mouse hippocampal neurons (36). We performed patch clamp experiments on rat primary hippocampal neurons. In immature neurons at DIV 2-4, IV relationships of 1000 M glycine-evoked currents clearly had an inwardly rectifying shape that resembled mostly those of transiently expressed GlyR ␣2 or ␣3 homomeric channels in HEK293T cells (Fig. 5B; compare with Fig. 3C). Messenger RNA coding for GlyR ␣3 was detected in primary hippocampal neurons, although at rather low levels compared with those corresponding to GlyR ␣2 and GlyR ␤ coding transcripts (Fig.  5A). The same holds true for more mature hippocampal neurons at DIV 15-25. However, there were significant differences in the shape of their IV relationships, which were much more V rev , reversal potential of the IV relationships of the indicated GlyR channels. V(I max ), membrane potential of the maximum outward current in the IV. ϪI ϩ150 mV /I Ϫ150mV , currents at ϩ150 mV normalized to currents at Ϫ150 mV. Numbers in parentheses indicate the number of experiments (n). For each data set, significant differences against linearity were calculated by a two-tailed t test of the data against V(I max ) ϭ 150 mV and ϪI ϩ150 mV /I Ϫ150mV ϭ 1. ND, not determined; NS, not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.  linear than those from immature neurons but still displayed substantial inward rectification (Fig. 5, B-E).

Discussion
GlyRs are ligand-gated chloride channels that contribute to synaptic signal transmission in the mammalian CNS. Currently, there are inconsistent data regarding biophysical properties of GlyRs due to heterogeneous heterologous expression systems. Here, we comparatively investigated the properties of glycine-evoked currents through recombinant GlyR channels using GlyR ␣1-3 splice variants as well as heteromeric GlyR ␣/␤ channels in the same heterologous expression system and using the same experimental setup. We studied glycine sensitivities after receptor desensitization and IV relationships of GlyR channels with different subunit compositions. The results reveal intriguing differences between homomeric GlyR ␣2/3 and ␣1 channels currents during desensitization with regard to IV rectification upon pronounced depolarization. Furthermore, our study identifies a single critical amino acid in transmembrane domain 2 that is responsible for voltage-dependent rectification of glycine-dependent currents in the desensitized state, which renders GlyR ␣1 and heteromeric GlyR ␣/␤ channels insensitive to voltage-dependent rectification of transmembrane chloride conductance during receptor desensitization. Apparent glycine sensitivity was higher in the desensitized state of GlyR ␣3L channels, and the Hill coefficient tended to be lower compared with the instantaneous responses to 1000 M glycine. Interestingly, this was the case not only for GlyR ␣3L channels but also for all other homomeric GlyR channels (except GlyR ␣2B). Although the EC 50 and n H values derived from concentration-response curves alone are not suited to directly distinguish between agonist binding and channel gating, due to the uncertain occupancy and cooperativity of the glycine binding sites as well as possible structural and kinetic M glycine. IV relationships were normalized to the currents at Ϫ150 mV. Insets, averaged and normalized time courses at Ϫ50 mV from 2 s before to 30 s after the start of glycine application. D, statistical analysis of the potentials at which the maximal outward currents were determined (V(I max )). The color code is similar to that in A-C. Note that only homomeric GlyR ␣2 and GlyR ␣3 channels display pronounced inward rectification. Error bars, S.E. differences in gating (37,38), our data suggest that the amino acid sequence homology between the different GlyR ␣ variants gives rise to similar desensitization properties at Ϫ50 mV. Indeed, the desensitization kinetics expressed as values were consistent between 5 and 10 s for each of the investigated homomeric GlyR ␣ channels. However, these data are not in good agreement with previous studies that reported that splicing of the long insert into GlyR ␣3 mRNA impacts on desensitization kinetics (12, 39 -41). Notably, desensitization of GlyR channels may be highly variable (20,41) and influenced by many parameters, such as receptor density (42), membrane potential (43), intracellular and extracellular pH (44,45), ion concentration (46,47), the membrane lipid composition, and phosphorylation of the intracellular TM3-TM4 domain or the N-glycosylation of the receptor N terminus (20). Technical features, such as slow or fast ligand application or the ligand con-centration (16,48), duration of ligand administration, or the recording mode may also account for these differences. Co-expression of the GlyR ␤ subunit did not change Hill coefficients of glycine-response curves and decay kinetics of GlyR ␣2and GlyR ␣3-containing channels, but the sensitivity of heteromeric channels was consistently increased by a factor of approximately 2 in comparison with the respective homomeric GlyR ␣2/3 variants, which is a new finding compared with recent publications that did not report such a tendency for ␣2-containing or ␣3-containing GlyRs (49 -51). Differences in the binding affinities of glycine between the sites formed by ␣/␣ or the ␣/␤ surfaces were previously resolved (52), possibly explaining these differences from previous studies.
We also investigated IV relationships of GlyR channels with different subunit compositions. The IV relationship of an ion channel can be regarded as a signature. It is an important FIGURE 4. An alanine residue within the GlyR␣3 pore is important for inward rectification. A, sequence comparison of the pore-lining regions of GlyR ␣1, GlyR ␣2, GlyR ␣3, and GlyR ␤ subunits (starting from position 243). Note that the sequences of all ␣ subunits are completely identical with the exception of one alanine/glycine at position 254. B, left, front (bottom) and top (top) view on homomeric GlyR ␣1 in the glycine-bound state (Protein Data Bank entry 1JAE). Gly 254 is located at the inner pore mouth and directed toward the channel pore. Right, structure homology model of a homomeric GlyR ␣3 channel. The side chain of Ala 254 is not directed toward the channel pore. C, averaged IV relationships of A254G mutants of GlyR ␣3K (top) and GlyR ␣3L (middle) upon desensitization and in response to 1000 M glycine. IV relationships were normalized to the currents at Ϫ150 mV. IV relationships of GlyR ␣3K wt and GlyR ␣3L wt are illustrated for comparison. Experiments were performed with the standard pipette solution. Insets, averaged and normalized time courses at Ϫ50 mV from 2 s before to 30 s after glycine application. Bottom, statistical analysis of V(I max ). Inward rectification of the A254G mutants was significantly reduced in comparison with WT. Error bars, S.E. parameter to characterize and identify an ion channel's biophysical properties or the subunit composition in vitro and in vivo. For GlyR channels, few data are available with respect to their IV relations, although glycine-induced currents through these channels are known to be voltage-dependent (21,48,53). However, many studies on IV relations of native or recombinant GlyR ␣1-containing channels were performed using different cellular expression systems, including neurons, HEK cells, and Xenopus oocytes (23, 24, 54 -56). In the present study, we describe for the first time IV relationships of homomeric and heteromeric GlyR channels over a wide voltage range from Ϫ150 to ϩ150 mV using the same HEK293T cellular expression system. The finding that IV relations of homomeric GlyR ␣2 and ␣3 channels change from linear to inwardly rectifying upon desensitization might be indicative for the changes of the structural configuration of the agonist binding sites, channel pore, linker sequences, and intracellular domains, altogether leading to the apparent changes in conductance (37). However, our data show that GlyR ␣3 inward rectification depends to a large extent on ion permeation because inward rectification of GlyR ␣3L was observed with both symmetrical and asymmetrical Cl Ϫ gradients, and the membrane potential of the onset of rectification (V(I max )) shifted to the same extent as the Nernst potential would predict. Thus, the inward rectification of desensitized GlyR ␣3L currents is not dependent on a given membrane potential but rather on a given chemical Cl Ϫ gradient across the membrane. Interestingly, a three-site/four-barrier rate model was previously suggested for GlyR ␣1 (25), and rectification of the single channel conductance can be changed by point mutations of the amino acids that putatively contribute to the barriers within the channel pore (26). We have found that inward rectification of a GlyR ␣3L mutant, A254G, was significantly reduced. Alanine at position 254 in the TM2 is located at the inner channel pore mouth of GlyR ␣3L and can be considered analogous to Gly 254 in GlyR ␣1, where it is located at position 2Ј in the pore-forming TM2 transmembrane helix (26,29). Substituting glycine with alanine actually deletes only a methyl group but does not introduce or delete charged or polar residues in a protein. It is therefore likely that, structurally, either the pore diameter is affected or the general positioning of TM2 is changed by the A254G mutation in GlyR ␣3L. However, our data do not favor one of these two possibilities, and the protein homology structure model of GlyR ␣3 (see Fig. 4) predicts that the methyl group of Ala 254 is not directly displayed toward the pore. Rather, mutational studies on GlyR ␣1 (57) have suggested that ion permeation is sterically regulated at the inner pore mouth between positions Ϫ3Ј and ϩ2Ј relative to TM2. It is furthermore accepted that the narrowest point in the GlyR pore is located at the inner mouth (29,58). Our results therefore suggest that the inwardly rectifying IV relationship of desensitized GlyR ␣3 channels is most likely determined by an asymmetrical energy barrier for Cl Ϫ permeation. Interestingly, gating and permeation of LGICs are closely linked, and desensitization was recently shown to be regulated by interactions between the TM2 and TM3 segments, affecting the lumen of the channel pores close to the intracellular side (18). Although position 254 (Gly 254 in ␣1, Ala 254 in ␣2 and ␣3) was not directly investigated in that study, it is close to the other positions (e.g. Gly 256 in ␣1) (18) that are involved in TM2-TM3 coupling and receptor desensitization. Thus, Ala 254 in GlyR ␣2 and ␣3 homomeric channels determines rectification of the IV relationship in the desensitized state, which is most likely due to an asymmetrical energy barrier for Cl Ϫ permeation. Our study also showed that inward rectification only occurs with homomeric GlyR ␣2 and ␣3 channels, consistent with the presence of an alanine at position 254. In contrast, IV relationships of homomeric GlyR ␣1 channels or heteromeric GlyR ␣2/␤ or GlyR ␣3/␤ channels were rather linear in the desensitized state. The inwardly rectifying IV relationship can therefore be regarded as a signature of homomeric GlyR ␣2 or ␣3 channels, which is useful for the discrimination of these GlyR channels from homomeric GlyR ␣1 or heteromeric GlyR ␣2 and ␣3 channels that contain the ␤ subunit. Indeed, primary hippocampal neurons investigated at DIV 2-25 were shown here to display GlyR responses with rather inwardly rectifying IV relationships upon desensitization, suggesting that homomeric GlyR ␣2 and/or ␣3 channels were expressed at these developmental states. Our RT-PCR analysis provides evidence that supports this conclusion because GlyR ␣2 was preponderantly expressed between DIV 3 and DIV 24 (Fig. 5A). The finding that mRNA coding for GlyR ␤ was also detected actually also supports recent evidence for the lack of GlyR ␤ protein expression in hippocampal neurons (36). Furthermore, these results indicate that the identified voltage-dependent rectification of homomeric GlyR ␣2/3 can be used as physio-molecular signature to identify these channels at the single cell level.
In vivo, GlyR ␣2 and ␣3 are expressed in different brain regions, including the neocortex and hippocampus, in both post-and presynaptic neuronal compartments (2,4,8,59,60). The intracellular chloride concentration of most mature neurons in the CNS is relatively low, shifting the onset of GlyR rectification in the physiological range of membrane potentials. Thus, postsynaptic plasma membrane homomeric GlyR ␣2 or GlyR ␣3 channels would restrain chloride conductance at depolarized membrane potentials (e.g. during a series of action potentials and resulting massive membrane depolarization) if glutamatergic synapses are active. However, at presynaptic terminals of glutamatergic synapses, homomeric GlyR ␣3L was shown to be expressed in presynaptic vesicles with the ligandbinding domain facing the lumen of vesicles (4). Because the vesicular membrane potential would be much more positive compared with the plasma membrane potential, ranging between ϩ40 and ϩ80 mV (61), voltage-dependent rectification of GlyR-dependent currents can be particularly relevant with regard to the loading of vesicles with neurotransmitters because it would contribute to increasing ⌬pH. The glycine transporter GlyT1 is also expressed at glutamatergic terminals in the vesicular membrane (62), and vesicle loading with glutamate indeed depends on the vesicular chloride gradient (63). Thus, vesicular GlyRs would be permanently desensitized (due to GlyT1-dependent glycine fill) and restrain chloride-facilitated glutamate loading of presynaptic vesicles, due to voltagedependent rectification.
Altogether, our study reveals that inward rectification of IV relationships of desensitized homomeric GlyR ␣2 and ␣3 is a physio-molecular signature of these channels and a peculiar biophysical property that restrains chloride conductance at depolarized membrane potentials or in compartments with a low chloride concentration, which can be particularly relevant for maladaptive forms of neuronal plasticity in neuropsychiatric disorders, such as autism (5), and epilepsy (3,4).