Studies of γ2 subunit-deficient mouse mutants have shown that the γ2 subunit of the GABAA receptor (GABAAR) is necessary for the postsynaptic clustering of the GABAARs and for the maintenance of GABAAR clusters at GABAergic synapses (Essrich et al. 1998; Schweizer et al. 2003). The γ2–/– mouse mutant shows a severe deficit in GABAergic synaptic transmission and dies soon after birth (Günther et al. 1995). GABAARs play a morphogenic role during embryonic development (Rudolph and Mohler 2004; Vicini and Ortinski 2004). Thus, some of the observed phenotypes in these and other mutant mice might result from developmental alterations, while the absence of phenotype might be because of compensatory mechanisms. RNA interference (RNAi, Dykxhoorn et al. 2003; Huppi et al. 2005) is a simpler alternative to the gene knockout technology that can also overcome some of the limitations inherent to the use of mouse mutants. In this study, we have used γ2 RNAi to study GABAAR clustering in loss-of-function experiments, both in neuronal cultures and in the intact brain after in utero electroporation. Our results support the notion that the γ2 subunit is necessary for the postsynaptic clustering and maintenance of GABAARs and gephyrin (a postsynaptic scaffolding protein that is present at inhibitory GABAergic and glycinergic synapses). More interesting, because we revealed it with RNAi technology but it had not been previously observed with the γ2 mouse mutants, is the observation that the disruption of the postsynaptic clustering of GABAARs in pyramidal neurons leads to decreased presynaptic GABAergic innervation of these neurons. The results are consistent with the notion that postsynaptic GABAAR clustering plays a role in the formation and/or stabilization of the presynaptic GABAergic contacts.
We have used RNA interference (RNAi) to knock down the expression of the γ2 subunit of the GABAA receptors (GABAARs) in pyramidal neurons in culture and in the intact brain. Two hairpin small interference RNAs (shRNAs) for the γ2 subunit, one targeting the coding region and the other one the 3′-untranslated region (UTR) of the γ2 mRNA, when introduced into cultured rat hippocampal pyramidal neurons, efficiently inhibited the synthesis of the GABAA receptor γ2 subunit and the clustering of other GABAAR subunits and gephyrin in these cells. More significantly, this effect was accompanied by a reduction of the GABAergic innervation that these neurons received. In contrast, the γ2 shRNAs had no effect on the clustering of postsynaptic α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors, postsynaptic density protein 95 (PSD-95) or presynaptic glutamatergic innervation. A γ2-enhanced green fluorescent protein (EGFP) subunit construct, whose mRNA did not contain the 3′-UTR targeted by γ2 RNAi, rescued both the postsynaptic clustering of GABAARs and the GABAergic innervation. Decreased GABAAR clustering and GABAergic innervation of pyramidal neurons in the post-natal rat cerebral cortex was also observed after in utero transfection of these neurons with the γ2 shRNAs. The results indicate that the postsynaptic clustering of GABAARs in pyramidal neurons is involved in the stabilization of the presynaptic GABAergic contacts.
axon initial segment
coding region with three point mutations
enhanced green fluorescent protein
γ-aminobutyric acid type-A receptor
glutamic acid decarboxylase
mouse monoclonal antibody
postsynaptic density protein 95
hairpin small interference RNAs
3′-untranslated region with three point mutations
synaptic vesicle GABA transporter
synaptic vesicle glutamate transporter-1
Materials and methods
All the animal protocols have been approved by the Institutional Animal Care and Use Committee and followed the National Institutes of Health guidelines.
All the anti-GABAAR antibodies were raised in our laboratory: The rabbit and guinea pig anti-rat γ2 GABAAR subunit antibodies were raised to amino acids 1–15. The rabbit anti-rat α2 GABAAR subunit antibody was raised to amino acids 417–423. The mouse monoclonal (mAb) anti-β2/3 GABAAR antibody (62–3G1) was raised to the affinity-purified GABAAR (De Blas et al. 1988; Vitorica et al. 1988). It recognizes an N-terminal epitope that is common to the rat β2 and β3 subunits but is not present in the β1 subunit (Ewert et al. 1992). These anti-GABAAR antibodies have been thoroughly characterized and used in several studies (De Blas et al. 1988; Vitorica et al. 1988; Miralles et al. 1999; Christie et al. 2002a,b; Christie and De Blas 2003; Charych et al. 2004a,b). The mouse monoclonal anti-postsynaptic density protein 95 (PSD-95) was from Upstate Biotechnology (Lake Placid, NY, USA), guinea pig anti-synaptic vesicle glutamate transporter-1 (vGlut1) and guinea pig anti-synaptic vesicle GABA transporter (vGAT) were from Chemicon (Temecula, CA, USA). The rabbit anti-GluR2/3 was a gift of Dr Robert J. Wenthold (NIDCD, Bethesda), the mouse mAb to SV2 was a gift of Dr Kathleen M. Buckley (Harvard Medical School, Boston), and the sheep anti-glutamic acid decarboxylase (GAD) was a gift of Dr Irwin J. Kopin (NINDS, Bethesda).
Hippocampal cultures were prepared by the method of Goslin et al. (1998) as described elsewhere (Christie et al. 2002a,b). Briefly, dissociated neurons from embryonic day 18 Sprague–Dawley rat hippocampi were plated at a density (10 000–20 000 cells per 18-mm diameter circular coverslip) and maintained in glial cell conditioned medium for 16–19 days. These cultures contained 90–95% pyramidal cells and 5–10% interneurons.
Generation of the small hairpin RNAs
Four small hairpin RNAs (shRNAs) constructs, based on the mU6pro vector (Yu et al. 2002; Bai et al. 2003), were made. The first shRNA (γ2 CR) targeted a sequence in the coding region (CR) of the rat γ2 subunit (nucleotides 131–155: GenBank accession no. L08497). A DNA oligonucleotide encoding both arms of the shRNA was annealed with the corresponding antisense DNA and the double-strand DNA was ligated between the BbsI and XbaI sites of the mU6pro vector. The transcription of this DNA generated the γ2 CR shRNA (Fig. 1a). The antisense strand of the shRNA perfectly matched the target mRNA, while the sense strand included a mismatch near the middle (Fig. 1a, lower-case nucleotide) to facilitate DNA sequencing. The second shRNA (γ2 UTR) targeted a sequence of the 3′-untranslated region (UTR) of the rat γ2 subunit (nucleotides 1467–1491); The third and fourth shRNA (γ2 CR3m and γ2 UTR3m) were used as control shRNAs for the first and second shRNAS, respectively, by introducing three point mutations in the sense and antisense arms of the corresponding shRNA (Fig. 1a, nucleotides in red color). The four shRNAs contained a three-nucleotide ACA loop, a G at the 5′ end and a UUUU sequence at the 3′ end. Lower-case nucleotides at the 5′ end and the 3′ end are from the transcription of the mU6 vector, not from the target sequence. The γ2 CR and γ2 UTR were designed to target both γ2S and γ2L mRNAs (Whiting et al. 1990).
Transfection of the shRNA in hippocampal cultured neurons
Cultured hippocampal neurons (9-day-old) were transfected with shRNA using the CalPhos Mammalian transfection kit (BD Bioscience, San Jose, CA, USA) following the instructions of the manufacturer. The transfection cocktail also included p-enhanced green fluorescent protein (EGFP)-N1 plasmid (Clontech, Palo Alto, CA, USA) to identify transfected cells by EGFP fluorescence. Other constructs based on the pEGFP-N1 plasmid (γ2-EGFP, α2-EGFP and β3-EGFP) made in our laboratory were also used for transfection. These constructs were used for the expression of GABAAR subunits, each tagged at the C-terminus with EGFP. The quality of the constructs was assessed by DNA sequencing and their expression in HEK293 cells and cultured hippocampal neurons was confirmed with subunit-specific antibodies and EGFP fluorescence. Two micrograms of each plasmid were used for the transfection of each culture in 18-mm2 coverslips. Seven to 10 days after transfection (16–19 days in culture) neurons were subjected to fluorescence immunocytochemistry.
Immunofluorescence of hippocampal cultures
Immunofluorescence of hippocampal cultures was carried out as described elsewhere (Christie et al. 2002a,b). Briefly, hippocampal cultures were incubated in 4% paraformaldehyde and 4% sucrose in phosphate-buffered saline (PBS) for 15 min at room temperature. Permeabilization was carried out with 0.25% Triton X-100 in PBS for 10 min at room temperature. Cells were incubated with 5% donkey serum in PBS for 30 min at room temperature followed by incubation with a mixture of primary antibodies in 0.25% Triton X-100 PBS for 2 h at room temperature. After washing, the cells were incubated with a mixture of fluorophore-labeled secondary antibodies (anti-species-specific IgG) raised in donkey and conjugated to Texas Red or AMCA (Jackson Immuno-Research, West Grove, PA, USA) for 1 h at room temperature. After washing, the coverslips were mounted with ProLong anti-fade mounting solution (Molecular Probes, Eugene, OR, USA).
In utero electroporation
With this technique, the DNA is injected in the lateral ventricle and the cells lining the ventricle are transfected by electroporation. The transfected neurons migrate into their final destination in the various layers of the cerebral cortex where they express the transfected protein or shRNA during several weeks after birth. Embryonic rat brain gene transfer by using in utero electroporation was carried out as described by Bai et al. (2003). Briefly, pregnant Wistar rats at 13-day gestation were anesthetized with 100 mg/kg ketamine-HCl, 10 mg/kg xylazine (intraperitoneally) and a laparotomy was performed. The uterine horns were gently pulled out and 1–3 μL of a sterile mixture of plasmids (1.5 μg/μL for each γ2 shRNA and 0.5 μg/μL for pLZRS-CA-gapEGFP, gifts from Drs A. Okada and S. K. McConnell, Stanford University, Stanford, or pCAGGS-DsRed, a gift from A. Nishiyama, University of Connecticut, Storrs) and Fast Green (2 mg/mL; Sigma, St Louis. MO, USA) were microinjected by pressure with a picospritzer through the uterine wall into the lateral ventricles of the embryos with a sterile glass capillary pipette. Electroporation was carried out by a brief (1–2-ms) discharge of a 500-μF capacitor charged to 50–100 V with a power supply. The voltage pulse was discharged with a pair of sterile gold/copper alloy oval plates (1 × 0.5 cm) after gently pinching the head of each embryo through the uterus. After electroporation, the uterus was returned to the body cavity and the incision was closed sewing it up with sterile surgical suture. The pups were killed 14–21 days after birth.
Immunocytochemistry of rat brain sections
Fourteen- to 21-day-old Wistar rats were anesthetized (80 mg/kg ketamine-HCl, 8 mg/kg xylazine, 2 mg/kg acepromacine maleate) and perfused through the ascending aorta with fixative consisting of 4% paraformaldehyde in 0.1 m phosphate buffer (PB), pH 7.4. The frozen brain was sliced in coronal sections (25-μm thick) with a freezing microtome. Free-floating sections were incubated at 4°C for 40 h with affinity-purified rabbit anti-γ2 antibody (1 : 100) and guinea pig anti-vGAT antibody (1: 2000) in 0.3% Triton X-100, 0.1 m PB, pH 7.4. The washed tissue sections were incubated with fluorescence-labeled secondary antibodies (Alexa 647-labeled goat anti-rabbit IgG and Alexa 555-labeled goat anti-guinea pig IgG from Molecular Probes. Sections were washed and mounted on gelatin-coated glass slides with ProLong Gold anti-fade mounting solution (Molecular Probes).
Image acquisition and analysis
For hippocampal cultures, images were collected using a 60X pan-fluor objective on a Nikon Eclipse T300 microscope with a Sensys KAF 1401E CCD camera, driven by IPLab 3.0 (Scanalytics, Fairfax, VA, USA) acquisition software. The images were analyzed with PhotoShop 4.01 (Adobe). Brightness and contrast were adjusted, the image was changed from 16 bits/channel to 8 bits/channel (1315 × 1035 pixel resolution), sharpened using the unsharp mask tool (settings: amount = 125%, radius = 1.5 pixel, threshold = 0 level), color was added to each channel and the images were merged for color co-localization. Fluorescent images in figures are presented before subtraction of the diffuse background fluorescent signal in dendrites. For rat brain sections, images were acquired on a Leica TCS SP2 laser confocal microscope using a HCXPL Apo 100X oil CS objective lens and a pinhole set at 1 Airy unit. For image processing and quantitative analysis, ImageJ software (NIH) was used.
Quantification of synaptic clusters and GAD+ boutons
For quantification of cluster density in cultured pyramidal cells, the maximum intensities of the fluorophore channels were normalized and the background fluorescence of each channel seen in the dendrites was substracted. For each combination of antibodies, three independent immunofluorescent experiments were performed. A total of 35–50 dendritic fields of randomly selected dendrites from 20 to 30 pyramidal neurons (7–10 neurons per experiment) were analyzed. Each measurement was taken from a 50-μm long dendritic segment (with an average width of 2 μm). Density values were calculated as number of clusters per 100 µm2 of dendritic surface. The number of clusters analyzed for each shRNA was in the range of 364–817. In the cases where clustering was highly inhibited by shRNAs, the total number of clusters ranged from 54 to 180. For quantifying the number of neurons that received GABAergic innervation, for each shRNA, 29–33 transfected and 42–52 non-transfected pyramidal neurons were randomly selected from three individual experiments (six coverslips), and the number of neurons that were contacted by any GAD+ presynaptic terminal was recorded. The densities of the GAD+ boutons were calculated as the number of boutons per cell.
To quantify γ2 subunit GABAAR clusters and presynaptic vGAT+ boutons on the cell surface of the pyramidal cells in the intact rat brain cerebral cortex, a paracircular line was drawn along the surface of the soma of the pyramidal neurons. The surface of the soma was identified in the transfected neurons that expressed EGFP by the edge of the cell fluorescence. The surface of the cell soma was also identified by increasing the background contrast of the vGAT or γ2 GABAAR immunofluorescence channels. The soma of the non-transfected pyramidal neurons was revealed by the relative absence of immunofluorescence compared with the surrounding neuropil. In transfected cells, the surface of the soma determined by this method was identical to the surface of the soma determined by EGFP fluorescence, thus validating both methods for determining the surface of the soma. The density of the γ2 GABAAR subunit clusters and vGAT+ terminals on the surface of the neuronal soma was determined by counting all the γ2 clusters and vGAT+ terminals localized within 1 μm on each side of the paracircular line marking the surface of the soma. Ten to 12 pyramidal neurons of each class (see Results) and 10 optical sections (0.5-μm thick each) for each neuron were analyzed (258–584 γ2 clusters and 553–1227 vGAT+ terminals per class were counted). Quantification in pyramidal neurons transfected with shRNA was compared with (i) that of non-transfected neighboring pyramidal neurons of similar size located in the same layer, and to (ii) that of pyramidal neurons in the same cortical layer from another animal transfected only with EGFP (with no shRNA). Values were averaged per cell and one-way anova Tukey's test was used for statistical analysis.
Knock-down of the γ2 GABAAR subunit by RNAi inhibits the clustering of GABAARs and gephyrin
A shRNA targeting the coding region of γ2 mRNA (γ2 CR) led to a large reduction of GABAAR cluster density (Fig. 1b) as shown by immunofluorescence with an anti-γ2 Ab (Fig. 1bi) or an anti-α2 Ab (Fig. 1biii). Transfected cells in this and other experiments were identified by EGFP fluorescence (green color, Fig. 1b, bii, biv). Compare in Fig. 1(b), the difference in GABAAR cluster density between the dendrites of the transfected neurons (Fig. 1b, green color), where few clusters are observed, with a dendrite of a non-transfected cell (Fig. 1b, arrow), which contains many GABAAR clusters. The density of GABAAR clusters (mean ± SEM) in the transfected neurons was 15.5 ± 1.8% (for γ2) and 19.1 ± 2.4% (for α2) of that of non-transfected neurons (Fig. 2e). Arrowheads in Fig. 1 show GABAAR clusters. Similarly, cells transfected with the shRNA targeting the 3′-UTR of the γ2 mRNA (γ2 UTR) also show a large reduction of GABAAR clusters (Fig. 1d). The transfected cells had 11.4 ± 1.7% (for γ2) and 16.2 ± 1.9% (for α2) of the corresponding cluster density of the non-transfected neurons (Fig. 2f). In contrast, the neurons transfected with shRNAs each having three point mutations (γ2 CR3m and γ2 UTR3m, respectively), showed no effect on GABAAR cluster density, as compared with non-transfected cells (Figs 1c and e, and Figs 2e and f). Note that the fluorescence intensity of the remaining GABAAR clusters (arrowheads) after γ2 RNAi is considerably lower than in control neurons (Figs 1bi and biii vs. ci and ciii, or di and diii vs. ei and eiii, respectively).
The decreased expression of the γ2 subunit resulting from γ2 RNAi also led to a drastic reduction in β2/3 GABAAR subunit clusters in the transfected cell (Fig. 2a, green color) compared with the dendrites of non-transfected neurons (Fig. 2a, arrow). Nevertheless, the β2/3 reduction was less pronounced than that of γ2 and α2 clusters. In neurons transfected with shRNA γ2 CR (Fig. 2a) or γ2 UTR (not shown), the β2/3 cluster density was reduced to 27.8 ± 3.2% (Figs 2ai and e) and 28.8 ± 3.4% (Fig. 2f), respectively. Some β2/3 clusters (Fig. 2ai, open arrowheads) remained in the neurons transfected with γ2 shRNA even when these clusters showed no clear γ2 immunoreactivity (Fig. 2aiii, open arrowheads) indicating that some of the β2/3 GABAAR subunit clusters might not contain γ2. In these GABAARs, γ1 or γ3 might substitute for γ2 (Baer et al. 1999). In contrast, the neurons transfected with the mutated shRNAs showed no significant changes in γ2 or β2/3 receptor clusters (Figs 2b, d, e and f). Note the high degree of co-localization of the γ2 with α2 clusters (Figs 1ci–civ and ei–eiv, filled arrowheads) and with β2/3 clusters (Figs 2bi–biv, filled arrowheads), respectively. This is expected as these subunits frequently co-assemble in the same pentameric (α2β2/3γ2) GABAAR.
Gephyrin is a postsynaptic scaffold protein present in many GABAergic and glycinergic synapses (Essrich et al. 1998; Kneussel et al. 1999; Sassoe-Pognetto et al. 2000; Brünig et al. 2002b). In hippocampal cultures, gephyrin forms clusters co-localized with GABAAR clusters (Craig et al. 1996; Rao et al. 2000; Christie et al. 2002a,b). Figure 2(c, e and f) shows that γ2 RNAi blocks gephyrin clustering. The neurons transfected with γ2 CR (Fig. 2e) or γ2 UTR shRNAs (green color, Fig. 2c) and Fig. 2(f) show a highly reduced density of gephyrin clusters (Fig. 2ci, open arrowheads) compared with that of non-transfected neurons or neurons transfected with γ2 UTR3m shRNA (Fig. 2di, filled arrowheads). The density of gephyrin clusters in the transfected neurons was 21.8 ± 1.9% (for γ2 CR) and 15.9 ± 1.8% (for γ2 UTR) of the cluster density in non-transfected neurons, respectively, as shown in Fig. 2(e and f). After γ2 RNAi, some gephyrin clusters did not show co-localizing γ2 clusters, indicating that some gephyrin clusters are not associated with GABAARs that contain the γ2 subunit (Fig. 2ci vs. Fig. 2ciii, open arrowheads). As indicated above, they might contain γ1 or γ3 instead. Neurons transfected with the mutated shRNAs (green color, Fig. 2d) showed no reduction of gephyrin or γ2 clusters (Fig. 2di and diii, respectively) compared with non-transfected neurons as shown in Fig. 2(e and f). Gephyrin and γ2 clusters showed a high degree of co-localization (Figs 2di–div, filled arrowheads).
In hippocampal cultures the knock-down of the γ2 GABAAR subunit leads to reduced GABAergic innervation of the targeted pyramidal neurons
We and others have used GAD immunostaining to identity presynaptic GABAergic terminals in neuronal cultures (Craig et al. 1996; Essrich et al. 1998; Rao et al. 2000; Brünig et al. 2002b; Christie et al. 2002b; Lévi et al. 2004). Nevertheless, we confirmed that in our cultures (Figs 3a, ai–aiv, filled arrowheads) all the GAD+ boutons contained the GABAergic synaptic vesicle marker vGAT (vesicular GABA transporter) and the general synaptic vesicle marker SV2, which is present in both GABAergic and glutamatergic presynaptic terminals. Some SV2 clusters (Fig. 3aii, open arrowhead) that do not co-localize with GAD or vGAT, instead co-localize with the presynaptic glutamatergic marker vGlut1 (not shown). Moreover, we have shown that, in these cultures, the presynaptic GAD+ boutons that are apposed to postsynaptic GABAARs have actively recycling synaptic vesicles (Christie et al. 2002b). The presynaptic GAD+ boutons that are apposed to the postsynaptic dendritic GABAAR clusters are varicose enlargements of the axon(s) that form ‘en passant synapses’. The GAD+ axons can be followed to their origin at the soma of the GABAergic interneurons (not shown).
Fig. 3(b and d) shows that pyramidal cells transfected (green) with γ2 CR or γ2 UTR shRNAs have a parallel reduction of both γ2 clusters (Figs 3biii and diii) and the GABAergic innervation of these neurons by GAD-containing interneurons (Figs 3bi and di) when compared with non-transfected pyramidal neurons (Figs 3b and d, arrows). Decreased GABAergic innervation was manifested as decreased proportion of transfected pyramidal cells receiving GABAergic innervation, and as decreased density of GAD+ boutons contacting the soma and dendrites of the pyramidal cells as compared with the non-transfected pyramidal cells (Figs 3f and g).
In pyramidal cells transfected either with γ2 CR or γ2 UTR shRNAs, the average densities of GAD+ boutons contacting the transfected pyramidal cells were 34.3 ± 3.4 and 27.8 ± 2.6%, respectively, of the average density of GAD+ boutons contacting non-transfected pyramidal cells (Fig. 3f). Moreover, the proportion of these neurons that showed GABAergic innervation (GAD+ boutons) was 41 ± 3 and 38 ± 4%, respectively, of the proportion of the non-transfected pyramidal cells that received GABAergic innervation (Fig. 3g).
Pyramidal cells transfected with the mutated γ2 shRNAs (Figs 3c and e, green color) showed no reduction in γ2 clusters or GAD+ terminals, or in the proportion of pyramidal neurons receiving GAD+ terminals, compared with non-transfected pyramidal cells (Figs 3f and g). Note the high degree of apposition in the cultures of the presynaptic GAD+ terminals and postsynaptic γ2 clusters (arrowheads, Figs 3ci–civ and ei–eiv).
Exogenous γ2-EGFP rescues both GABAAR clustering and the GABAergic innervation blocked by γ2 RNAi
As indicated above, the shRNA that targets the γ2 UTR induced a large decrease in both GABAAR cluster density and GABAergic innervation of the neurons. Co-transfection of γ2-EGFP with γ2 UTR shRNA rescued to a large extent the formation of γ2 GABAAR clusters (from 11.4 ± 1.7 to 84.5 ± 3.3%, p = 0.001, one-way anova Tukey test, Fig. 4g) and GABAergic innervation (from 27.8 ± 4.2 to 88.1 ± 4.8%, p = 0.001, Fig. 4h) as shown by the presence of presynaptic GAD+ boutons that co-localized with (were apposed to) postsynaptic γ2 GABAAR clusters (Figs 4ai, aiii and aiv, arrowheads). The formation of synaptic γ2-EGFP fusion protein clusters was revealed by the presence of EGFP fluorescence at GABAergic synapses (Fig. 4aii, arrowheads). The γ2-EGFP clusters co-localized with the γ2 clusters revealed an anti-γ2 antibody (Fig. 4ai, arrowheads). This antibody recognized the N-terminus of both endogenous γ2 and exogenous γ2-EGFP. The density of γ2 GABAAR clusters (84.5 ± 3.3%) and GAD+ boutons (88.1 ± 4.8%) in the γ2-EGFP rescued neurons was not significantly different (p > 0.05, one-way anova Tukey test) from that of the neurons that were co-transfected with the mutated shRNA γ2 UTR3m (96.4 ± 5.1 and 91.5 ± 3.4%, respectively, Figs 4bi–biv, arrowheads, and Figs 4g and h) or from that of non-transfected neurons (100%). In these experiments, the γ2-EGFP constructs (and β3-EGFP and α2-EGFP, see below) did not contain the 3′-UTRs that are present in the endogenous γ2, β3 or α2, as EGFP was fused to the C-terminus of each subunit. Therefore, the shRNA γ2 UTR targeted the endogenous γ2 mRNA but not the transfected γ2-EGFP mRNA. The γ2-EGFP also rescued the clustering of endogenous α2 and β2/3 GABAARs, which co-localized with γ2 GABAAR clusters and γ2-EGFP clusters (not shown) at GABAergic synapses.
These results strongly suggest that the γ2-EGFP co-assembles with endogenous GABAAR subunits forming GABAARs that target the endogenous α2 and β subunits to GABAergic synapses. Others have shown that EGFP-tagged GABAAR subunits co-assemble with endogenous GABAAR subunits to form functional GABAARs (Bueno et al. 1998; Connor et al. 1998;Kittler et al. 2000; Alldred et al. 2005).
The α2-EGFP or β3-EGFP, when transfected into cultured pyramidal neurons, formed clusters that co-localized with endogenous GABAAR subunits at GABAergic synapses. However, contrary to γ2-EGFP, when the β3-EGFP (Fig. 4c) or α2-EGFP (Fig. 4e) were co-expressed with γ2 UTR shRNA they did not rescue GABAAR clustering (Figs 4ci and ei) or GABAergic innervation (Figs 4ciii and eiii), as the values were not significantly different (p > 0.05, one-way anova Tukey test) from that of neurons transfected with γ2 UTR shRNA only, as quantification in Fig. 4(g and h) shows (21.1 ± 3.1 or 17.9 ± 3.3 vs. 11.4 ± 1.7% for γ2 GABAAR clusters and 32.5 ± 2.3 or 31.2 ± 3.2% vs. 27.8 ± 4.0% for GAD+ terminals, respectively). The neurons transfected with the mutated γ2 UTR3m shRNA (Figs 4d and f) showed that β3-EGFP (Fig. 4dii) and α2-EGFP (Fig. 4fii) formed clusters (arrowheads) that co-localized with the endogenous γ2 (Fig. 4di), α2 (Fig. 4fi) and β2/3 subunit (not shown) and with presynaptic GAD+ terminals (Figs 4diii and fiii, arrowheads). Note that the anti-α2 subunit antibody used in Fig. 4(ei and fi) recognized the C-terminus of the endogenous α2, but not the α2-EGFP fusion protein, which is EGFP-tagged at the C-terminus of α2. Thus, in Fig. 4(ei and fi), only the endogenous α2 but not α2-EGFP was recognized by this antibody.
These results are also consistent with the aforementioned notion that the exogenous α2-EGFP or β3-EGFP co-assemble with the endogenous γ2 and other endogenous α and β subunits forming GABAAR clusters at GABAergic synapses. However, in the absence of the endogenous γ2 (after γ2 RNAi), the α2-EGFP and β3-EGFP no longer formed GABAAR clusters at GABAergic synapses. These experiments also support the notion that the γ2 subunit is essential for the synaptic clustering of the GABAARs. They are also consistent with a recent report showing that, in cultured neurons from the γ2–/– mouse mutant, the synaptic clustering of GABAARs could be rescued by a different EGFP-γ2 construct protein (Alldred et al. 2005) and by the transgenic expression of γ2S or γ2L (Baer et al. 2000).
In hippocampal cultures the knock-down of the γ2 GABAAR subunit does not affect AMPA receptor clustering or the glutamatergic innervation of the targeted pyramidal cells
Transfection of pyramidal neurons with shRNA γ2 UTR (Fig. 5a, green color) highly decreased the density of GABAAR clusters (Fig. 5aiii). However, in the same cell, there was no effect on the density of GluR2/3 clusters (Fig. 5ai, arrowheads) or PSD-95 clusters (106.3 ± 3.4 and 95.5 ± 4.6%, respectively) when compared with non-transfected cells (100%) or pyramidal cells transfected with the mutated shRNA γ2 UTR3m (102.5 ± 4.3 and 95.5 ± 4.3%, respectively) as shown in Fig. 5(b and bi arrowheads, and Fig. 5h, p > 0.05, one-way anova Tukey test). Note that the latter shows normal γ2 GABAAR clustering (Fig. 5biii, arrows). Inhibition of GABAAR clustering but no effect on α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor or PSD-95 clustering (109.2 ± 4.2 and 106.5 ± 3.9%, respectively) was also observed in pyramidal neurons transfected with the shRNA γ2 CR (Fig. 5g). No effect on either GABAAR or AMPA receptor or PSD-95 clustering was observed in pyramidal neurons transfected with the mutated γ2 CR3m shRNA (97.6 ± 3.8 and 109.4 ± 5.6%, respectively) as shown in Fig. 5(g). Similarly, experiments using an antibody to the glutamatergic presynaptic marker vGlut1 (Figs 5c–f, red color, arrowheads) showed that shRNA γ2 CR (Fig. 5c) or γ2 UTR (Fig. 5e) did not significantly affect glutamatergic presynaptic innervation of the pyramidal cells (109.4 ± 5.6% or 105.2 ± 3.5%, respectively) as compared with non-transfected cells (100%) or with pyramidal cells transfected with the mutated γ2 CR3m (Fig. 5d) or γ2 UTR3m (Fig. 5f) shRNAs (98.3 ± 6.5 or 94.5 ± 5.2%, respectively; p > 0.05, one-way anova Tukey test) as shown in Fig. 5(g and h).
All the controls put together indicate that the γ2 RNAi effects are specific: (i) the target sequences of the γ2 shRNAs are specific for the γ2 mRNA, as shown by sequence databases' search; (ii) the same inhibitory effects were obtained with two shRNAs, one targeting the coding region and the other one targeting the 3′-UTR of the γ2 mRNA; (iii) no effect was observed by transfecting only with the p-EGFP-N1 plasmid; (iv) the RNAi effects were abolished by introducing three point mutations in each of the shRNAs; (v) the γ2 RNAi specifically reduced GABAAR clustering and GABAergic innervation, but not the clustering of AMPA receptors, PSD-95 or glutamatergic innervation, thus showing that γ2 RNAi does not induce a generalized protein shutdown; and (vi) the clustering of GABAARs and the GABAergic innervation that were reduced by the γ2 RNAi that targeted the 3′-UTR of the endogenous γ2 mRNA, were rescued by a γ2-EGFP mRNA devoided of the target-3′-UTR, but were not rescued by α2-EGFP or β3-EGFP.
In the intact brain, the knock-down of the γ2 subunit in a subpopulation of pyramidal neurons leads to a reduction of both GABAAR clusters and GABAergic terminals innervating the targeted neurons
We tested whether the γ2 RNAi effects observed in hippocampal cultures were extended to the intact brain. Transgenic expression of γ2 shRNAs in the intact brain was achieved after in utero electroporation of the rat embryos (E13). The post-natal (P14 and P21) brains showed many transfected pyramidal neurons distributed thoughout all layers of the cerebral cortex as shown by EGFP fluorescence (Fig. 6g). Neurons transfected with γ2 UTR shRNA (green fluorescent neuron 1, Figs 6a–c) showed decreased density of GABAAR clusters and vGAT+ boutons on their surface when compared with neighboring EGFP– neurons (neuron 2, Figs 6a–c) or with neurons from other animals from the same layer transfected only with EGFP but not with γ2 UTR shRNA (neuron 3, Figs 6d–f) or non-transfected neighboring neurons (neuron 4, Figs 6d–f). Quantification (Fig. 6h) shows that the neurons transfected with γ2 UTR shRNA have reduced density of γ2 GABAAR clusters and vGAT+ boutons on their surface (47 ± 3 and 76 ± 3%, respectively, p < 0.001) compared with controls (neuron 3, Figs 6d–f). Similar results were obtained when γ2 CR shRNA was used for in utero electroporation (Fig. 6i). In this case, the density of γ2 GABAAR clusters and vGAT+ boutons in the surface of the transfected pyramidal neurons (class 1) were 47 ± 2 and 71 ± 2% of control (class 3), respectively (p < 0.001). These results indicate that, in the intact brain, γ2 RNAi decreased both γ2 GABAAR clusters and GABAergic innervations of the shRNA expressing neurons. Figure 6(h and i) also shows that there is no significant difference in the density of GABAAR clusters and GABAergic innervation among various classes of control neurons (classes 2, 3 and 4, one-way anova Tukey test).
The results also show that γ2 RNAi had a larger inhibiting effect on the GABAAR cluster density (47%) than on the GAD+ bouton density (71–76%). This phenomenon was also observed in the cultures as shown in Fig. 4(g)[11–15% for γ2 GABAARs as shown in Fig. 2 (e and f) vs. 28–34% for GAD+ boutons as shown in Fig. 3f]. The results also show that the γ2 RNAi exerted a larger effect in the cultured neurons than in the intact brain. We do not know yet whether this difference is because of the higher difficulty when using brain tissue (compared with the cultures) in differentiating between the synapses that are located on the soma of the pyramidal cells subjected to γ2 RNAi and the synapses from the neuropil that are very close to but not on the soma of these cells. The latter synapses would not be subjected to γ2 RNAi, thus masking the true extent of the γ2 RNAi effect on the GABAAR clustering and GABAergic innervation.
The γ2 RNAi inhibited the synaptic clustering of the α2 and β2/3 GABAAR subunits and gephyrin. By using a very different experimental approach, our γ2 RNAi experiments also led to the notion that the γ2 subunit is essential for the clustering and cluster maintenance of the GABAARs and gephyrin at GABAergic synapses. This hypothesis was originally derived from studying the γ2–/– mouse mutant (Essrich et al. 1998; Schweizer et al. 2003), but it had not been corroborated by an independent approach. This hypothesis is also consistent with electrophysiology and EM immunocytochemistry experiments with wild-type animals. These experiments show that many of the GABAARs that contain the γ2 subunit are synaptically localized and are responsible for the synaptic phasic inhibition, while the GABAARs that do not contain the γ2 subunit (i.e. the ones containing δ instead of γ2) are non-synaptic and responsible for the tonic GABAergic inhibition (Nusser et al. 1998; Brickley et al. 1999; Christie et al. 2002a; Nusser and Mody 2002; Wei et al. 2003; Yeung et al. 2003). Gephyrin also plays a role in the synaptic clustering of some GABAARs, although the extent of which is less established (Kneussel et al. 1999; Kneussel and Betz 2000; Kneussel et al. 2001; Lévi et al. 2004; Lüscher and Keller 2004).
The most significant observation derived from our studies is that, in both the intact brain and in neuronal cultures, γ2 RNAi led, not only to a reduction of GABAAR clusters in the transfected pyramidal neurons, but also to a reduction of the GABAergic innervation that these neurons received. This was shown by both a decreased proportion of pyramidal cells receiving GABAergic innervation and by a decreased density of GABAergic terminals contacting the pyramidal neuron expressing γ2 shRNA. This result was unexpected as it has been reported that the principal neurons of γ2–/– mouse mutants showed no reduced GABAergic innervation, either in the intact brain or in neuronal cultures (Essrich et al. 1998; Schweizer et al. 2003). A possible explanation for this apparent discrepancy is that, in our RNAi experiments, mosaics are formed where only the neurons transfected with the γ2 shRNA have reduced density of GABAAR clusters. The γ2 shRNA transfected neurons would be in competitive disadvantage for GABAergic synapse formation and/or stabilization with their non-transfected neighbors, which fully express γ2 and have normal density of GABAAR clusters. However, in the γ2–/– knockout mouse, none of the neurons expresses γ2 or forms normal GABAAR clusters, so there is no competitive advantage or disadvantage of any neuron for maintaining GABAergic innervation. Moreover, in the case of the γ2–/– knockout mouse, it is likely that the GABAergic presynaptic terminals are mismatched to postsynaptic clusters of NMDA receptors and PSD-95 (Rao et al. 2000), which might support the mismatched maintenance of the presynaptic GABAergic contacts.
Another possible and more general explanation has been given in other examples where RNAi produced a phenotype while the corresponding gene knockout mutant had no phenotype (Bai et al. 2003; Götz 2003). In gene knockouts, the absence of a phenotype is often because of the existence of redundant proteins or the compensatory expression of proteins having similar function. Compensatory protein expression might not be activated by γ2 RNAi. The RNAi allows acute knock-down of mRNAs during specific time periods rather than deleting the gene throughout the lifespan of the animal. Moreover, although RNAi can reduce the specific mRNA up to 80–90% (Dykxhoorn et al. 2003), and the translated protein by even more, some residual γ2 protein is expressed by the transfected cells (Figs 2e and f). This small level of mRNA and protein expression in combination with the timing of the RNAi might prevent the activation of the compensatory mechanisms. Another possibility is that there are species differences (i.e. our studies on rats vs. the studies on mouse mutants) in their response to the loss of function of the γ2 protein.
We and others have shown that, in hippocampal cultures, GABAARs and gephyrin form postsynaptic clusters at GABAergic synapses (Craig et al. 1996; Lévi et al. 1999; Brünig et al. 2002b; Christie et al. 2002a,b; Christie and De Blas 2003; Fritschy et al. 2003). We have also shown that, upon GABAergic innervation, large GABAAR and gephyrin clusters are formed at GABAergic synapses, while there is disappearance of small GABAAR and gephyrin clusters in the area surrounding the GABAergic synapses (Christie et al. 2002b). Thus, innervation by presynaptic GABAergic terminals affects both the location and size of GABAAR and gephyrin clusters by inducing the concentration of GABAARs at the postsynaptic membrane in apposition to the presynaptic GABAergic terminals. In this study, we have shown that the postsynaptic clustering of GABAARs affects the formation and/or stabilization of the presynaptic GABAergic contacts. Thus, the ability to form GABAAR clusters seems to be one of the factors involved in the stabilization of GABAergic innervation. This mechanism might not be unique to GABAergic synapses. In mature hippocampal cultures, the axon initial segment (AIS) of the pyramidal neurons is innervated by GABAergic terminals but not by glutamatergic terminals (Brünig et al. 2002a; Craig and Lichtman 2002; Christie and De Blas 2003). However, in young cultures there are glutamatergic terminals contacting the AIS of pyramidal neurons (Christie and De Blas 2003). We have hypothesized (Christie and De Blas 2003) that the withdrawal of the presynaptic glutamatergic innervation from the AIS might be as a result of the documented absence of postsynaptic glutamate receptor clusters and PSD-95 clusters in the AIS of these neurons. However, GABAergic innervation of the AIS would be stabilized by the formation of large GABAAR clusters and gephyrin clusters in the AIS of these neurons.
As indicated above, the postsynaptic clustering of GABAARs is essential for the postsynaptic clustering of gephyrin. It is also likely that GABAAR clustering is essential for the clustering of some of the other proteins that are present at the postsynaptic GABAergic complex, including cell surface recognition molecules. Thus, it is conceivable that disruption of the postsynaptic clustering of GABAARs by γ2 RNAi also disrupts the postsynaptic clustering of cell-recognition molecules, which in turn disrupts the interaction of the latter with presynaptic cell recognition partners, thus leading to decreased GABAergic innervation. The cell recognition molecule neuroligin 2 concentrates postsynaptically at GABAergic synapses, interacts with presynaptic neurexins, and both neuroligin 2 and neurexins are involved in the transynaptic differentiation and accumulation of pre- and postsynaptic elements at GABAergic synapses (Graf et al. 2004; Varoqueaux et al. 2004; Chih et al. 2005). Moreover, in cultured hippocampal pyramidal cells, the knock-down of postsynaptic neuroligin 2 by RNAi, leads to reduced density of presynaptic GABAergic contacts on these cells (Chih et al. 2005). This effect is similar to what we have observed by the knock-down of the postsynaptic γ2 GABAAR subunit by γ2 RNAi. Nevertheless, it is not known yet whether there is a relationship between the postsynaptic GABAAR clustering and the postsynaptic co-localization of neuroligin 2 at GABAergic synapses. Other cell surface recognition molecules, like some ephrins and ephrin receptors, interact with GRIP proteins (Bruckner et al. 1999; Contractor et al. 2002). GRIP proteins form clusters at GABAergic and glutamatergic synapses, often co-localizing with GABAAR clusters (Charych et al. 2004b; Kittler et al. 2004; Li et al. 2005). Thus, in addition to neurexins and neuroligin 2, it is conceivable that some ephrins, ephrin receptors and/or other cell recognition molecules might also be involved in the stabilization of GABAergic synaptic contacts. Alternatively, the effect of GABAAR clustering on the stabilization of presynaptic GABAergic contacts could be mediated by soluble growth factors whose release by the postsynaptic neuron concentrates at the sites where GABAARs cluster. In any of these models (involving cell-surface recognition molecules or growth factors), the disruption of postsynaptic GABAAR clustering would interfere with the formation and/or stabilization of GABAergic synapses. Synaptic activity does not play a major role in the clustering of postsynaptic GABAARs or the maintenance of the presynaptic GABAergic contacts. Thus, chronic exposure of the cultures to bicuculline or tetrodotoxin has no significant effect on GABAAR clustering or GABAergic innervation (Craig et al. 1994; Craig 1998; Rao et al. 2000; Studler et al. 2002).
Our results are consistent with the hypothesis that GABAAR clustering is one of the factors regulating GABAergic synapse formation and/or stabilization.
We thank Dr Bih Y. Yang for the construction of the pEGFP-β3 subunit plasmids. This research was supported by the National Institute of Neurological Disorder and Stroke (Grants NS38752 and NS39287).