Preferential accumulation of GABA_A receptor γ2L, not γ2S, cytoplasmic loops at rat spinal cord inhibitory synapses

Methods

All experiments were carried out according to the guidelines laid down by the Landesamt für Arbeitsschutz, Gesundheit und technische Sicherheit Berlin (T0406/98). Rats were deeply anaesthetised with ether before sacrifice by decapitation.

RT and cDNA constructs

RNA was isolated from adult rat spinal cord using Trizol (Invitrogen Life Technologies, Carlsbad, CA, USA). Complementary DNA (cDNA) was obtained by reverse transcription (SuperscriptII, GibcoBRL, Bethesda, MD, USA) of RNA using oligo-dT primer. PCR conditions were: 30 cycles of 45 s at 95°C, 45 s at 58°C and 1 min at 72°C. For green fluorescent protein (GFP)-receptor loop chimeras (GFP::γ2S/L), PCR amplimers comprising the cytoplasmic loop were obtained using oligonucleotides (5′GGGGAATTCGCATTATTTTGTGAGCAACCGG3′ and 5′GGGGGATCCTCAGGAGTCCATTTTGGCAATGCG3′). After subcloning using EcoRI and BamHI into the mammalian expression vectors pEGFP-C1 and pDsRed2-C1 (Clontech, Palo Alto, CA, USA) we distinguished between GABA_AR γ2S and γ2L loops with the following primers (S/L: 5′CAAGGTCTCCTATGTCACAGC3′ and 5′AAGGCGGTAGGGAAGAAGATC3′, yielding 377 and 401 bp, respectively). Green fluorescent protein-tagged gephyrin (GeC2,6) was isolated by RT-PCR as previously described (Meier & Grantyn, 2004). Final constructs were verified by DNA sequencing. Oligonucleotides used for the RT-PCR analysis of the GABA_AR subunit expression in DIV12 SCNs have been described elsewhere (Meier et al. 2002).

Site-directed mutagenesis

Site-directed mutagenesis was performed using the GeneEditor system (Promega, Madison, WI, USA). The GABA_AR γ2L loop was subcloned into Bluescript using EcoRI and BamHI. The 5′-phosphorylated oligonucleotide 5′CTTCGGATGTTTGCCTTCAAGG3′ served to exchange the Ser343 residue against alanine (S343A). The resulting γ2L–S343A loop was re-cloned into pEGFP-C1 and verified by DNA sequencing.

Cell culture and transfection

Spinal cord neurones (SCNs) of embryonic day 14 Wistar rats were prepared as previously described (Meier et al. 2002) and maintained on top of spinal cord glia cells in Neurobasal medium (GibcoBRL) supplemented with B27 (GibcoBRL) and 1% fetal calf serum (Brewer & Cotman, 1989). Transfection was carried out after 12 days in vitro (DIV) using Effectene reagent, as previously described (Meier & Grantyn, 2004). COS-7 cells were transfected using FuGENE6 reagent (Roche Diagnostics, Mannheim, Germany). After 72 h of protein expression, African green monkey kidney fibroplast cells (COS-7 line) cells were processed for immunocytochemistry.

Antibodies

The mouse monoclonal antibodies (mAbs) recognized gephyrin (mAb7a, 1:200, Pfeiffer et al. 1984; Alexis Biochemicals, San Diego, CA, USA) and DsRed (1:100, Clontech). Polyclonal antibodies (pAbs) were guinea-pig anti-vesicular inhibitory amino acid transporter (VIAAT; 1:800, Dumoulin et al. 1999; Chemicon, Temecula, CA, USA), guinea-pig anti-GABA_AR γ2 (1:4000, Fritschy & Mohler, 1995) and rabbit anti-GFP (1:300, Molecular Probes, Eugene, OR, USA]. For multiple labelling, mAbs and pAbs were combined. To ensure that labelling was specifically due to primary antibodies we replaced them with similarly diluted normal serum from the same species. Secondary antibodies were carboxymethyl-indocyanine (Cy3 and Cy5), fluorescein-isothiocyanate (FITC), coupled and purchased from Jackson ImmunoResearch Laboratories (West Grove, PA, USA), or AlexaFluor350, coupled and purchased from Molecular Probes.

Immunocytochemistry and quantification

Immunofluorescence was performed as previously described (Meier & Grantyn, 2004). Images were acquired using a 12-bit cooled CCD camera (Ch250, Photometrics, Tucson, AZ, USA) mounted on a standard epifluorescence microscope (objective 100×, Axiovert, Carl Zeiss, Jena, Germany). Appropriate filters (XF22, XF32, XF110-2 and XF136-2; Omega Optical Inc., Brattleboro, VT, USA) allowed the detection of FITC, Cy3, Cy5 or AlexaFluor350 fluorescence.

To distinguish between diffuse fluorescence signals and clusters, images were background subtracted within the linear range of an 8 bit grey scale. Fluorescence intensity profiles were obtained along GFP-expressing dendrites (NIH Image software). GFP distributes diffusely within the cytosol (Llopis et al. 1998). The standard deviation (s.d.) of the mean of this diffuse fluorescence signal was obtained and compared with the s.d. of GFP::γ2S and GFP::γ2L. The respective values are 9.1 ± 0.9 (GFP, n = 85), 15.1 ± 1.4 (GFP::γ2S, n = 51) and 35.8 ± 2.5 (GFP::γ2L, n = 58). This is the first indication that the fluorescence patterns obtained with GFP::γ2S and GFP::γ2L were significantly different (P < 0.001, Student's unpaired t test). The classification of a fluorescence signal as a cluster required that its maximal fluorescence above the mean value exceeded the s.d. of the diffuse GFP fluorescence 3 times (3σ criterion). In addition, a threshold was set at one-third of the maximal fluorescence of a presumed cluster. Classification as a cluster required the width of the fluorescence peak at the threshold to be in the range of 3 (0.33 μm) and 30 pixels (3.3 μm).

To determine colocalization, images were merged. For colocalization of two clusters of immunogens, we required that >50% of the pixels reflecting one immunogen should show colocalization with a coherent cluster of the corresponding other immunogen. The term ‘apposition’ is used when two immunogens are closely associated (distance <5 pixels and overlap <50%). Colocalized clusters were counted within a region of interest comprising the soma and 50 μm of proximal dendrites. A colocalization index was defined by evaluation of colocalization (apposition) of corresponding immunoreactivities and expressed as the percentage fraction of the number of colocalized clusters and the total number of clusters of a given immunoreactivity.

Since we could not determine the distribution of our expression constructs by cell surface staining of the antigens, we analysed the fluorescence distribution along a line delineating the outer perimeter of the COS-7 cells, i.e. the plasma membrane (e.g. Kins et al. 2000).

If not mentioned otherwise, values were obtained from 10–20 cells in 3 independent experiments and expressed as means ± s.e.m. The level of significance was determined by Student's unpaired t test and is indicated by one (P < 0.05), two (P < 0.01) or three (P < 0.001) asterisks.

Results

The large intracellular loop of GABA_AR γ2S accumulates ineffectively at inhibitory postsynaptic sites

GABA_AR γ2S and γ2L large intracellular loops were tagged with GFP to visualize their subcellular localization. With the resulting constructs (GFP::γ2S/L), SCNs were transfected on DIV12 and displayed GFP fluorescence after 16 h of protein expression. According to the approach to discriminate between diffuse and clustered fluorescence (Methods), GFP::γ2S was found diffusely distributed within the neuronal cytosol (Fig. 1Aa and B). Occasionally GFP::γ2S formed clusters (e.g. Fig. 1B, high-power view, arrow). VIAAT-immunostaining of inhibitory presynaptic terminals revealed that GFP::γ2S rarely accumulated opposite inhibitory terminals (Fig. 1B). The fraction of VIAAT-immunoreactive terminals that colocalized with GFP::γ2S was 14.4 ± 2.1% (Fig. 1C, left-hand, shaded column). To find out whether the occasional binding of GFP::γ2S to postsynaptic molecules was associated with the presence of endogenous gephyrin, we performed triple immunofluorescence staining, as shown in Fig. 1B. When GFP::γ2S clustered opposite VIAAT-immunoreactive terminals GFP::γ2S was likely to colocalize with gephyrin (Fig. 1B, high-power view, arrow). The fraction of postsynaptic GFP::γ2S that colocalized with gephyrin was 83.3 ± 16.7% (Fig. 1C, right-hand, hatched column). These experiments show that GABA_AR GFP::γ2S accumulates with low efficacy at gephyrin-containing postsynaptic sites.

The large intracellular loop of GABA_AR γ2L accumulates at inhibitory postsynaptic sites with significantly higher efficacy than γ2S

To find out whether DIV12 SCNs express the GABA_AR γ2L splice variant, RT-PCR analysis was performed. As shown in Fig. 2A, DIV12 SCNs expressed the GABA_AR subunits (target) α1–3, β2, β3, γ2S and γ2L. Upper bands represent the internal control β-actin, which was coamplified in the same PCR reaction tubes to visualize equal amounts of cDNA. Since DIV12 SCNs expressed both GABA_AR γ2S and γ2L splice variants, we examined the ability of GFP::γ2L to accumulate at inhibitory postsynaptic sites. As shown in Fig. 1Ab and B, GFP::γ2L also tended to distribute diffusely within the neuronal cytosol. However, in contrast to GFP::γ2S, GFP::γ2L aggregates were easily distinguishable from the diffuse GFP signal (Figs 1Ab, 1Ac and 2B). VIAAT immunostaining revealed that GFP::γ2L aggregates were associated with inhibitory presynaptic terminals (Fig. 2B, high-power view, arrows). Of the examined VIAAT-immunoreactive boutons, 47.7 ± 3.4% colocalized with, or were apposed to, GFP::γ2L (Fig. 2C, left-hand, shaded column). This fraction was significantly (P < 0.001, Student's unpaired t test) higher than the respective value obtained for GFP::γ2S. To find out whether the postsynaptic accumulation of GFP::γ2L was associated with the presence of endogenous gephyrin, triple immunofluorescence staining was performed as shown in Fig. 2B. Of the synapses that displayed association of GFP::γ2L with VIAAT, 85.3 ± 4.8% also contained gephyrin (Fig. 2B, high-power view, arrows and Fig. 2C, right-hand, hatched column). To rule out the possibility that lack of colocalization between GFP::γ2S or GFP::γ2L and VIAAT was due to an absence of gephyrin in front of VIAAT-positive terminals, we determined the respective fractions of VIAAT terminals that displayed colocalization with gephyrin. We found that gephyrin accumulated equally at VIAAT sites in GFP::γ2S-transfected (87.0 ± 2.5%, n = 51) and GFP::γ2L-transfected neurones (91.3 ± 1.8%, n = 58). We therefore concluded that, in comparison with GFP::γ2S, GFP::γ2L has a significantly higher capacity to accumulate at gephyrin-containing postsynaptic sites.

Competition for postsynaptic accumulation between endogenous γ2-containing GABA_ARs and GABA_AR GFP::γ2L, but not GFP::γ2S

To ensure that the GFP-tagged GABA_AR γ2 loops can be used as surrogates for full-length GABA_ARs, we analysed whether the transfected loops competed with endogenous GABA_ARs for postsynaptic accumulation. SCNs were transfected with GFP::γ2S or GFP::γ2L, and triple labelling experiments were performed to visualize the GFP signal together with endogenous γ2-containing GABA_ARs and gephyrin. If the transfected GABA_AR γ2 loops and endogenous GABA_AR γ2 shared a common postsynaptic binding site, GFP::γ2L should compete with endogenous GABA_AR γ2. Transfection of GFP::γ2S had no effect on the distribution of endogenous γ2-containing GABA_ARs (Fig. 3Aa–c, arrows). In nontransfected and GFP::γ2S-transfected neurones, 80.9 ± 2.6 and 79.1 ± 3.4% of gephyrin clusters, respectively, were immunoreactive for endogenous γ2-containing GABA_ARs (Fig. 3C). However, whenever GFP::γ2L was associated with endogenous gephyrin (Fig. 3Ba–c, dendrite of a transfected neurone, crossed arrows) endogenous γ2-containing GABA_ARs were almost entirely excluded from postsynaptic loci. In GFP::γ2L-transfected neurones, the fraction of gephyrin clusters that contained endogenous γ2-containing GABA_ARs decreased significantly to a value of 40.8 ± 3.6% (Fig. 3C). This decrease was due to a competition between endogenous γ2-containing GABA_ARs and transfected GFP::γ2L, since the fraction of GFP::γ2L-associated gephyrin clusters that colocalized with endogenous γ2-containing GABA_ARs was only 10.5 ± 3.8% (Fig. 3C, right-hand, hatched column). As a control, in the same field of view, the association of endogenous γ2-containing GABA_ARs with gephyrin in nontransfected neurones remained unaltered (Fig. 3Ba–c, arrows, and C). Therefore, it is likely that GFP::γ2L and endogenous γ2-containing GABA_ARs share a common binding site in the postsynaptic apparatus of inhibitory synapses.

PKC phosphorylation of Ser343 facilitates postsynaptic accumulation of GFP::γ2L

It has previously been shown that PKC activation facilitates postsynaptic accumulation of GABA_ARs in embryonic CNS brain slices (Meier et al. 2003). We therefore explored the possibility that postsynaptic accumulation of GABA_AR GFP::γ2S or GFP::γ2L could be enhanced by increasing the PKC activity. Short-term treatment with the phorbol ester phorbol-12,13-dibutyrate (PDBu) (1 μm, 30 min before fixation) had no effect on the distribution of GFP::γ2S (Figs 4Aa and 5D). Again, we assume that a diffuse GFP::γ2S signal reflects the absence of postsynaptic accumulation or binding to components within the postsynaptic protein network. However, PDBu treatment (Figs 4Ba and 5D) caused a significant increase in the fraction of VIAAT-immunoreactive terminals colocalizing with GFP::γ2L (76.5 ± 4.6%) versus controls (47.7 ± 3.4%). The PDBu effect on the GFP::γ2L distribution could be antagonized by coapplication of the PKC blocker chelerythrine chloride (CC; Figs 4Bb and 5D). In that case, the level of colocalization (8.3 ± 3.3%) dropped to values comparable with the postsynaptic accumulation of GFP::γ2S. Since CC alone (Figs 4Bc and 5D) reduced postsynaptic GFP::γ2L accumulation to levels (9.6 ± 4.5%) comparable with those of GFP::γ2S, it seems that a basal PKC phosphorylation underlies the higher capacity of GFP::γ2L to accumulate at inhibitory postsynaptic sites.

The additional peptide LLRMFSFK within GFP::γ2L contains the potential PKC phosphorylation site Ser343 (Moss et al. 1992). The fact that GFP::γ2S gathered ineffectively and in a PKC-insensitive manner suggests that Ser343 might be a target of PKC phosphorylation. To examine the role of this insert, the Ser343 residue was substituted by alanine, and DIV12 SCNs were transfected with the resulting expression construct GFP::γ2L–S343A. Both in the absence (Fig. 5A) and in the presence (Fig. 5C) of PDBu, GFP::γ2L–S343A was found to be diffusely distributed within the neuronal cytosol (colocalization with VIAAT: 11.7 ± 4.3% in controls versus 10.1 ± 2.9% after PDBu treatment, Fig. 5B and D). Because GFP::γ2L-S343A behaved like GFP::γ2S we would like to propose that Ser343 phosphorylation contributes to a conformation that facilitates γ2L-subunit-mediated postsynaptic GABA_AR anchoring.

To finally find out whether PDBu treatment would allow for association of gephyrin with γ2L in non-neuronal cells, we cotransfected COS-7 cells with GFP-tagged gephyrin and DsRed2-tagged γ2L. To rule out optical filter cross-talk, the distribution of both proteins was visualized by indirect immunofluorescence using Alexa350- and Cy5-coupled secondary antibodies (Fig. 6A, GeC2,6::GFP and DsRed2::γ2L in false colours green and red, respectively (online)). As expected, under control conditions DsRed2::γ2L colocalized with a minor fraction (Fig. 6B, 12.4 ± 5.8%) of intracellular gephyrin aggregates. To our surprise, however, short-term PDBu treatment (30 min, 1 μm) resulted in a significant increase of this fraction to a value of 53.7 ± 6.7% (Fig. 6B). In 74% of cotransfected cells (Fig. 6C), PDBu treatment stimulated the appearance of a distinct population of small gephyrin spots (Fig. 6Ab, arrows) and, in addition, provoked the spatial association of gephyrin with the presumed plasma membrane (Fig. 6Ab, arrowheads). The projection area of the small gephyrin spots was ∼6-fold smaller than that of the usually observed gephyrin aggregates (0.18 ± 0.01 versus 1.1 ± 0.1 μm², Fig. 6D). In PDBu-treated cells, when gephyrin occurred in small aggregates or when it decorated the plasma membrane, it colocalized with DsRed2::γ2L (Fig. 6B–D). This PDBu effect was not observed when CC was present, except that small gephyrin aggregates were still found in 46% of cotransfected cells (Fig. 6Ac, arrows, and D).

Discussion

The use of protein fragments instead of full-length proteins is a common and widely used experimental model to study the molecular information contained in such protein subdomains. For example, it was shown that incorporation of the gephyrin binding sequence of the GlyR β-subunit into the NMDA receptor NR1 or the GABA_AR β1-subunit enabled the respective proteins to interact with gephyrin in non-neuronal cells (Meyer et al. 1995; Kins et al. 1999). Also using this approach, different gephyrin splice variants were compared with regard to their capacity to bind the cytoplasmic loop of the GlyR β-subunit (Meier et al. 2000). Gephyrin isoforms restraining GlyR binding were identified and shown to exclude GlyRs from postsynaptic anchoring opposite GABA-releasing terminals (Meier & Grantyn, 2004).

Here we chose the large cytoplasmic loops of the GABA_AR γ2S- and γ2L-subunits: (i) to characterize their individual binding efficacy to the postsynaptic proteome; and (ii) to rule out side-effects, such as phosphorylation-induced alterations in the cycling rate of full-length GABA_ARs between cell surface and intracellular compartments (Kittler et al. 2000). We obtained evidence that the GFP-tagged large cytoplasmic loop of the GABA_AR γ2L-subunit accumulates at inhibitory postsynaptic sites with significantly higher efficacy than the GABA_AR γ2S loop. This difference was even more pronounced after short-term PKC activation. The fact that postsynaptic GABA_AR γ2L loop accumulation was sensitive to the PKC antagonist chelerythrine chloride even in the absence of PDBu suggests that a basal level of PKC activity may be required for postsynaptic GABA_AR localization.

That PKC-mediated phosphorylation has profound effects on inhibitory synaptic transmission has been shown before (Brandon et al. 1999, 2000, 2002; Henneberger et al. 2002; Meier et al. 2003), but it is new that elimination of the PKC phosphorylation site Ser343 in γ2L loops affects the cellular distribution of this loop to resemble that of γ2S. Therefore, we have evidence to propose that the relative amount of γ2S and γ2L isoforms and the phosphorylation status of Ser343 are critical for the control of GABA_AR access to inhibitory postsynaptic sites.

Postsynaptic GABA_AR stabilization requires gephyrin, and this conclusion has been confirmed in GABA_AR γ2-gene-depletion experiments (Essrich et al. 1998; Schweizer et al. 2003). However, a direct interaction of purified GABA_AR preparations with gephyrin has not been demonstrated (Meyer et al. 1995; Kannenberg et al. 1997). Our finding that in non-neuronal cells, a widely used assay system (Kirsch et al. 1995; Kins et al. 1999, 2000; Kneussel et al. 1999, 2000; Grosskreutz et al. 2001), recruitment of GABA_AR γ2L loops to intracellular gephyrin aggregates required PKC activation provides a possible explanation for this failure. Nonetheless, a number of questions still need to be answered. (i) Is the association of the GABA_AR γ2L loop with gephyrin related to phosphorylation-induced changes in gephyrin? (ii) Does recruitment of gephyrin to the plasma membrane depend on the phosphorylation of gephyrin, or the GABA_AR γ2L loop, or both? (iii) Is the association of gephyrin and GABA_AR γ2L a prerequisite for their targeting to the plasma membrane?

It has been shown that gephyrin is a phosphoprotein (Langosch et al. 1992). However, it seems unlikely that the phosphorylation status of gephyrin contributes to the regulation of GABA_AR binding, since the kinase that phosphorylates gephyrin is endogenous to highly purified GlyR preparations (Langosch et al. 1992). As for question (ii), a recent study on intracellular GlyR trafficking identified GlyR binding to gephyrin as an important determinant for their membrane delivery (Hanus et al. 2004). It is conceivable that the association of the GABA_AR γ2L-subunit with gephyrin, which is facilitated upon PKC activation, plays a similar role (Studler et al. 2004) by providing a permissive conformation of gephyrin or the GABA_AR γ2L-subunit. (iii) Collybistin and GABARAP also contribute to the regulation of receptor delivery to the postsynaptic membrane (Kins et al. 2000; Kittler et al. 2001a). Therefore, other proteins, including yet unknown phosphoproteins, need still to be considered in the GABA_AR trafficking to and anchoring at the postsynaptic plasma membrane.

Our results may be considered provocative because there is previous evidence that under some experimental conditions GABA_AR γ2S and γ2L can substitute for each other (Homanics et al. 1999; Baer et al. 2000). However, considering that those conclusions were drawn from rescue experiments on the genetic background of GABA_AR γ2-gene deficiency (Homanics et al. 1999; Baer et al. 2000), it may well be that GABA_AR γ2S-subunits were only functional because there was no GABA_AR γ2L to compete with GABA_AR γ2S for gephyrin binding sites.

Our experiments revealed that in comparison with the γ2S loop the isolated GABA_AR γ2L loop displayed a higher efficacy of postsynaptic accumulation. However, this does not preclude a certain contribution of GABA_AR γ2S to postsynaptic anchoring. In our experiments with neuronal cultures the transfected neurones continued to express endogenous GABA_AR γ2S- and γ2L-subunits, which then competed with the transfected loops. This might have resulted in a competitive disadvantage for the GABA_AR γ2S loops.

The relative significance of GABA_AR γ2L may in addition be age dependent. It is known that during development GABA_AR γ2L is up-regulated to become the predominant isoform in mature GABAergic synapses (Wang & Burt, 1991; Poulter et al. 1993; Roberts & Kellogg, 2000). This underscores the physiological relevance of the additional LLRMFSFK sequence in mature CNS structures. However, at earlier developmental stages, or under specific experimental conditions, postsynaptic GABA_AR accumulation may be enabled by other GABA_AR subunits. For example, the GABA_AR γ3-subunit was shown to substitute for the missing GABA_AR γ2 (Baer et al. 1999). Triller and colleagues demonstrated that in hippocampal cultures GABA_AR β3-subunits formed postsynaptic clusters at gephyrin-containing inhibitory synapses before γ2-subunits were recruited to inhibitory synapses (Danglot et al. 2003). In embryonic collicular slices, which still lacked the GABA_AR γ2L isoform, postsynaptic GABA_AR stabilization was also enhanced after activation of PKC (Meier et al. 2003).

This issue remains attractive because there is increasing evidence that a disturbance of the γ2S–γ2L balance correlates with disease. During the past 30 years, extensive efforts have been made to gain more information about the mechanisms underlying schizophrenia. Multiple alterations at the molecular level have been described, including a decrease in the expression level of the GABA transporter GAT1, the GABA synthesizing enzyme GAD67 and the tissue GABA content (Volk & Lewis, 2002; Volk et al. 2002; Lewis et al. 2004). Today, the deficient GABA neurotransmission within the prefrontal cortex is considered to contribute to schizophrenia. In this context, it is particularly interesting that schizophrenic persons displayed altered expression levels of various GABA_AR subunits, including a predominance of GABA_AR γ2L over γ2S transcript levels in the prefrontal cortex (Huntsman et al. 1998) and an increase in the expression of GABA_AR α2 (Volk et al. 2002). An altered relationship between GABA release and postsynaptic GABA_AR density may reflect a compensatory process aimed at keeping GABAergic transmission at a safe level. In any case, these clinical findings, along with our results and developmental studies (Wang & Burt, 1991; Poulter et al. 1993; Roberts & Kellogg, 2000), underline the important role of the GABA_AR γ2L-subunit in activity-related synaptic reorganization.