We have studied the effects of GABAergic innervation on the clustering of GABAA receptors (GABAARs) in cultured hippocampal neurons. In the absence of GABAergic innervation, pyramidal cells form small (0.36 ± 0.01 μm diameter) GABAAR clusters at their surface in the dendrites and soma. When receiving GABAergic innervation from glutamic acid decarboxylase-containing interneurons, pyramidal cells form large (1.62 ± 0.08 μm breadth) GABAAR clusters at GABAergic synapses. This is accompanied by a disappearance of the small GABAAR clusters in the local area surrounding each GABAergic synapse. Although the large synaptic GABAAR clusters of any neuron contained all GABAAR subunits and isoforms expressed by that neuron, the small clusters not localized at GABAergic synapses showed significant heterogeneity in subunit and isoform composition. Another difference between large GABAergic and small non-GABAergic GABAAR clusters was that a significant proportion of the latter was juxtaposed to postsynaptic markers of glutamatergic synapses such as PSD-95 and AMPA receptor GluR1 subunit. The densities of both the glutamate receptor-associated and non-associated small GABAAR clusters were decreased in areas surrounding GABAergic synapses. However, no effect on the density or distribution of glutamate receptor clusters was observed. The results suggest that there are local signals generated at GABAergic synapses that induce both assembly of large synaptic GABAAR clusters at the synapse and disappearance of the small GABAAR clusters in the surrounding area. In the absence of GABAergic innervation, weaker GABAAR-clustering signals, generated at glutamatergic synapses, induce the formation of small postsynaptic GABAAR clusters that remain juxtaposed to glutamate receptors at glutamatergic synapses.
- GABAA receptor
- subunit isoform
- synapse formation
- neuron culture
- glutamate receptor
Low-density hippocampal cultures, in combination with fluorescence immunocytochemistry, have proven very useful for studying the clustering of GABAAreceptors (GABAARs) in individual GABAergic synapses (Killisch et al., 1991; Craig et al., 1996). Observations from hippocampal and spinal cultures have revealed the colocalization of postsynaptic GABAAR clusters (Craig et al., 1994;Levi et al., 1999) with presynaptic GABAergic boutons that contain glutamic acid decarboxylase (GAD). In addition, cultured neurons showed GABAAR clusters that did not colocalize with GAD boutons (Kannenberg et al., 1999; Levi et al., 1999; Scotti and Reuter, 2001).
Despite these aforementioned studies, there has not been a systematic study on the characteristics of the GAD-related and GAD-independent GABAAR clusters. Moreover, in these studies, the possible heterogeneity of the GABAA subunit or isoform composition in the receptor clusters has not been addressed. This is an important issue in light of observations that in the intact brain and retina some GABAergic synapses and puncta show selectivity for certain α subunit-isoform-containing GABAARs (Fritschy et al., 1992; Koulen et al., 1996; Nusser et al., 1996a; Fletcher et al., 1998; Nyiri et al., 2001).
In the present study, we have used low-density hippocampal cultures in combination with triple-label immunofluorescence to examine (1) the organizing effects that the presynaptic GABAergic innervation exerts on synaptic and extrasynaptic GABAAR clustering in hippocampal pyramidal neurons, (2) the GABAAR subunit and isoform expression in individual cells in culture, (3) whether there is selectivity in the subunit and isoform composition in the various synaptic and extrasynaptic GABAAR clusters, and (4) the relationship between GABAAR clusters and glutamate receptor clusters.
MATERIALS AND METHODS
Antibodies. The primary antibodies, guinea pig anti-α1 [1–15 amino acids (aa)], rabbit anti-α1 (1–15 aa), rabbit anti-α2 (417–423 aa), rabbit anti-α3 (1–13 aa), and rabbit anti-γ2 (1–15 aa), were raised and affinity purified in our laboratory against synthetic peptides made to unique extracellular epitopes (N terminus for α1, α3 and γ2, and C terminus for α2) of rat GABAAR subunits (Miralles et al., 1999). The monoclonal mouse anti-β2/3 (62–3G1) was raised in our laboratory to affinity-purified GABAAR (De Blas et al., 1988; Vitorica et al., 1988). This antibody recognizes an extracellular N-terminus epitope that is common to β2 and β3 subunits but is not present in β1 (Ewert et al., 1992). Antibodies to fusion proteins of the intracellular loops of β1, β2, and β3 were also raised in rabbits in our laboratory and affinity purified with purified intracellular loop of the respective isoform (Moreno et al., 1994; Li and De Blas, 1997). Subunit and isoform-specific antibodies made in several species in our laboratory have allowed us to study colocalization by triple-label immunofluorescence (see below). All antibodies to GABAAR subunits used in this study have been thoroughly characterized, and their specificities have been determined elsewhere (De Blas et al., 1988; Vitorica et al., 1988; Moreno et al., 1994; Miralles et al., 1999). Specificity tests of GABAAR antibodies included ELISA, immunoblotting, light microscopy immunocytochemistry, displacement of immunoreactivity in these assays by specific peptides, and subunit-specific staining in host-transfected cell lines. The specificity of some antibodies was also tested for the absence of immunoreactivity in knock-out mouse mutants. The monoclonal mouse anti-gephyrin (mAb 7a) was purchased from Cedarlane (Accurate Chemical and Scientific Corp., Westbury, NY). Rabbit anti-GluR1 was from Chemicon (Temecula, CA), mouse monoclonal anti-PSD-95 was from Upstate Biotechnology (Lake Placid, NY), and mouse monoclonal anti-SV2 was a gift of Dr. Kathleen M. Buckley (Harvard Medical School). Sheep anti-GAD (gift of I. Kopin), GAD 65-specific mouse monoclonal GAD6 (Developmental Studies Hybridoma Bank, University of Iowa), affinity-purified rabbit anti-GABA Transporter-1 (GAT-1) from DiaSorin (Stillwater, MN), and affinity-purified rabbit anti-synaptic vesicle GABA Transporter (VGAT) from Alpha Diagnostics International (San Antonio, TX) were used for identifying interneurons and GABAergic presynaptic processes. For the exo-endocytotic assay, a rabbit antibody to the lumenal N terminus (1–23 aa) of synaptotagmin-I (Syt-N) from StressGen Biotechnologies(Victoria, B.C., Canada) was used.
Low-density and micro-island hippocampal cultures. Hippocampal cultures were prepared as described by Banker and Goslin (1998). Briefly, embryonic day 18 Wistar rat pup hippocampi were dissected in HBSS, followed by treatment with 0.25% trypsin (Sigma, St. Louis, MO) and trituration using a fire-polished Pasteur pipette. Dissociated cells were centrifuged in HBSS for 2 min at 1500 rpm, and the pellet was resuspended in plating medium [10% horse serum (Invitrogen) in DMEM with 0.6% glucose and 26 mm NaHCO3]. The suspended cells were plated at a density of 5,000–10,000 cells per 18-mm-diameter circular coverslip treated with poly-l-lysine (Sigma). These cultures contained 90–95% pyramidal cells and 5–10% interneurons (Benson et al., 1994). At the aforementioned plating density, pyramidal cells receive limited GABAergic innervation. The coverslips with cell suspension media were then placed in 5% CO2 at 37°C for 3–4 hr to allow settling and attachment of cells. The coverslips with attached cells were then placed upside down in 60-mm-diameter Petri dishes containing a glia-conditioned medium and placed in a 5% CO2 atmosphere at 37°C. After 2–3 d, a final concentration of 5 × 10− 6 m cytosine arabinoside was added for 16 hr. Cultures were maintained by replacement of one-half volume of fresh N2-supplemented DMEM every 3–5 d for 19–22 d.
For micro-island cultures, cells were prepared as above and plated on 18 mm coverslips that had been prepared by a modified method described previously by Segal (1991). The coverslips were coated with 0.2% Agarose, dried overnight under UV light, then sprayed with a misted solution of 1% poly-l-lysine in 1 m borate buffer, pH 8.5, and dried again. This was finally followed by overnight incubation with 10% horse serum (Invitrogen) in DMEM with 0.6% glucose and 26 mm NaHCO3 before hippocampal culturing. The maintenance schedule of micro-island cultures was identical to that described above for other cultures.
Glial-conditioned medium was prepared as described by Banker and Goslin (1998). Briefly, astroglia were prepared by tryptic dissociation of postnatal day 0 (P0) rat cortex and plated at a density of 300,000 cells/5 ml of DMEM with 0.6% d-glucose, 26 mmNaHCO3, and 10% horse serum on poly-l-lysine-coated 35 mm Petri dishes. Glial cultures were maintained until they reached 80–100% confluence (10–14 d). Two days before hippocampal culture, media was fully exchanged with DMEM containing N2 supplement (Invitrogen), 1% d-glucose, 0.1% ovalbumin, 1 mm sodium pyruvate, and 26 mmsodium bicarbonate.
Fluorescence immunocytochemistry. Triple-label immunofluorescence detection of GABAAR subunit isoforms, gephyrin, and GAD was conducted by fixing 19–22 d in vitro (DIV) cultured hippocampal neurons in a PBS solution containing 4% paraformaldehyde and 4% sucrose for 15 min at room temperature followed by permeabilization with 0.25% Triton X-100 in PBS. Nonspecific antibody labeling was minimized by treatment with 5% donkey serum in PBS for 30 min at room temperature. Primary antibodies were diluted in 0.25% Triton X-100 PBS and then applied to coverslips, followed by incubation for 2 hr at room temperature. To ensure the specificity of the various GABAAR subunit isoform antibodies in the triple-label fluorescence immunocytochemistry, control conditions were examined in which each primary anti-GABAAR subunit antibody was incubated for 30 min with 20 μg/ml of the corresponding peptide sequence, before incubation with cultures. Specificity was also demonstrated from the data presented in which some receptor clusters were labeled with some subunit isoform-specific antibodies but not others. After incubation with primary antibodies, cultures were twice washed for 5 min with PBS, followed by application of secondary antibodies raised in donkey and conjugated to Texas Red, FITC, or aminomethylcoumarin fluorophores (1:150 dilution in 0.25% Triton X-100 PBS; Jackson Immunochemicals) for 1 hr at room temperature. The coverslips were again twice washed with PBS for 5 min each and mounted using Prolong anti-fade mounting solution (Molecular Probes, Eugene, OR). For surface labeling of receptor subunits, cultures were fixed as above, followed by antibody incubations and washes without Triton X-100.
Exo-endocytotic assay. An exo-endocytotic assay was performed for labeling the recycling synaptic vesicles in living neurons according to a modification of a previously described method (Matteoli et al., 1992; Kraszewski et al., 1995; Bacci et al., 2001). Briefly, 3- and 4-week-old neuronal cultures were incubated with a rabbit antibody to the N-terminal domain of Syt-N in N-2-supplemented DMEM in a 5% CO2 atmosphere or 15 mm HEPES-DMEM, pH 7.2, for 1 hr at 37°C. After incubation with the primary antibody, coverslips containing the cultured neurons were washed twice for 5 min with the same medium at 37°C. The neurons were fixed with PBS containing 4% paraformaldehyde and 4% sucrose for 15 min at room temperature followed by permeabilization with 0.25% Triton X-100 in PBS and incubation with sheep anti-GAD, and guinea pig anti-γ2 or mouse anti-gephyrin, for triple-label immunofluorescence, as described above. To ensure that Syt-N labeling at 37°C was caused by the exposure of the synaptotagmin N-terminal epitope by exo-endocytotic activity, neurons were incubated with the Syt-N antibody at 4°C in N-2-supplemented 15 mm HEPES-DMEM, pH 7.2. Coverslips were washed twice for 5 min at 4°C with the same medium, then fixed and incubated with guinea pig anti-γ2 and sheep anti-GAD antibodies, as indicated above.
Image acquisition and analysis. Images were collected using a 60× pan-fluor objective on a Nikon Eclipse T300 microscope with a Sensys KAF 1401E CCD camera, driven by IPLab 3.0 (Scanalytics, Fairfax, VA) acquisition software. Image files were then processed and merged for color colocalization figures using PhotoShop 4.01 (Adobe). Control slides in which one or more primary antibodies were omitted showed no spillover in the other two fluorescence channels. Random drift of the fluorescence signal of the sample between channels was controlled by alignment of all channels using triple-labeled fluorescent microspheres (0.1 and 0.4 μm diameter; Molecular Probes). Quantification of colocalized signal was performed by normalizing intensity data between fluorophore channels followed by the subtraction of background fluorescence signal seen in the dendrites. The two or three color channel images to be compared were merged, and GABAAR clusters were counted over a 50 μm section of dendritic shaft and compared for colocalization or juxtaposition to other clusters. To determine the mean and SE for each condition, a minimum of 15 measurements were made of randomly selected dendrites from pyramidal neurons that showed limited GABAergic innervation in different areas of the coverslip. For experiments requiring matched dendrites within the same neuron, two dendrites were selected on the basis of dendrite thickness, one with GABAergic innervation and one without. Quantitation of the density of large and small clusters along 50 μm segments was performed as above.
Measurements of cluster size, area, and average fluorescence intensity were performed using IPLabs 3.0 software. Twelve-bit images (4096 grayscale intensity levels) were segmented, on the basis of fluorescence intensity levels, to create a binary mask that maximized the number of clusters for analysis. For comparisons of small and large clusters, data were collected from different areas of the same neuron to eliminate bias between neuronal samples.
Two types of GABAAR clusters are present in hippocampal neurons that receive GABAergic innervation: GABAergic innervation induces the formation of large GABAAR clusters located at GABAergic synapses; small GABAAR clusters are not associated with GABAergic innervation
The expression and clustering of GABAAR subunits and gephyrin in relationship to GABAergic innervation were examined in 19–23 DIV cultured neurons by using triple-label immunofluorescence with combinations of GABAAR subunit isoform-specific antibodies in conjunction with an anti-gephyrin antibody and antibodies to GABAergic presynaptic markers such as GAD, VGAT, and GAT-1. We have included gephyrin in this study because it has been shown that in cultured hippocampal and spinal motor neurons gephyrin forms clusters that frequently colocalize with GABAAR clusters in GABAergic synapses (Craig et al., 1996; Levi et al., 1999). Gephyrin also colocalizes with GABAARs in the intact brain, retina, and spinal cord (Levi et al., 1999; Fischer et al., 2000; Sassoe-Pognetto et al., 2000). Additionally, it has been proposed that gephyrin is involved in the postsynaptic clustering of GABAARs (Essrich et al., 1998; Kneussel et al., 1999), although no direct binding of GABAARs to gephyrin has been demonstrated and other proteins might also be involved in GABAAR clustering (Knuesel et al., 1999, 2001; Wang et al., 1999; Fischer et al., 2000; Kneussel et al., 2000, 2001).
Two types of receptor clusters were observed with the variousGABA AR subunit-specific antibodies used in this study, such as α1, β2/3, γ2 (Fig.1A,D,G,H,J,K), α2, α3 (Fig.2 B,E), β1, and β2 (Fig.3 A,D): (1) large GABAAR clusters [1.62 ± 0.08 (SEM) μm breadth (range = 0.65–5.5 μm); 1.14 ± 0.09 μm2 area (range = 0.15–7.83 μm2); n = 136 clusters] that colocalized with GAD-containing boutons (Fig.1 A–C; Figs.1 D–L, 2A–F, 3A–F, arrows) at GABAergic synapses and (2) smaller GABAAR clusters [0.36 ± 0.01 μm diameter (range = 0.2–0.65 μm); 0.09 ± 0.01 μm2 area (range = 0.02–0.26 μm2); n = 197 clusters] that did not colocalize with the GAD-containing boutons (Figs.1 D–L, 2A–F, 3A–F, filled arrowheads). In addition to a greater size of the GABAAR clusters colocalizing with GABAergic contacts, there was a greater average fluorescence intensity (1565 ± 21 intensity units per pixel;n = 136) when compared with extrasynaptic clusters present in other areas of the same neuron (1178 ± 9 intensity units per pixel; n = 197 clusters; p < 0.001), indicating that the large clusters have a higher receptor density than the smaller clusters. Often, the largest postsynaptic GABAAR clusters of pyramidal neurons were composed of several smaller clusters (Fig. 1 D).
The GAD-containing endings also contained the synaptic vesicle GABA transporter VGAT (Fig.4 A–C) and the presynaptic membrane GABA transporter GAT-1 (data not shown). Thus, the existence of a complete set of presynaptic and postsynaptic GABAergic elements plus the demonstrated existence of functional GABAergic synapses in these cultures, as shown by electrophysiological techniques (Segal and Barker, 1984; Jensen et al., 1999), indicated that the observed GABAergic innervation of pyramidal cells by the axonal endings from interneurons established functional synapses at the points where the presynaptic and postsynaptic elements concentrated and converged. We further demonstrated that this is the case by showing that the studied GABAergic contacts have exo-endocytotic activity of synaptic vesicles. For this purpose we have used an antibody (Syt-N) to the lumenal N-terminus domain of the synaptic vesicle protein synaptotagmin (Fig. 4 E). In this assay (Matteoli et al., 1992; Bacci et al., 2001) at 37°C, only the functional presynaptic terminals undergoing exo-endocytosis of synaptic vesicles became labeled with the Syt-N antibody. This labeling does not occur when the cultures are exposed to the antibody at 4°C, a temperature at which synaptic vesicle fusion with the presynaptic membrane did not occur (Fig. 4 H). Additionally, when the antibody recognizing the cytoplasmic protein gephyrin was incubated together with the Syt-N antibody at 37°C, there was labeling of active synapses by the Syt-N antibodies without detectable labeling of gephyrin clusters (data not shown), indicating that cytoplasmic antigens that do not become exposed to the cell surface are not immunolabeled. Therefore, Figure 4, D–F, shows that labeling of recycling synaptic vesicles occurs in the GAD-containing terminals and colocalizes with the large postsynaptic GABAAR clusters.
Almost every neuron examined (175 of 177) expressed clusters of both the γ2 subunit-containing GABAAR and gephyrin (Fig.1 A–F). Moreover, 94.7% of all γ2 subunit-containing GABAAR clusters colocalized with gephyrin clusters of identical size and shape, and 90.0% of gephyrin clusters colocalized with γ2 (Fig.1 D,E; see Table 2). The size, distribution, and density of clusters per cell varied depending on the neuron type and total amount of innervation present. The two types of clusters were found in pyramidal cells and interneurons, although pyramidal neurons typically had a higher density of clusters than did GAD-positive interneurons.
Surface labeling of GABAAR clusters under nonpermeabilizing conditions with antibodies raised against extracellular N-terminus epitopes of α1, α3, β2/3, and γ2 showed that both the synaptic clusters and the small clusters were surface expressed. Under the same nonpermeabilizing conditions, antibodies to the intracellular proteins gephyrin and GAD or to the cytoplasmic epitopes of GABAAR showed no immunolabeling. Intracellular proteins and epitopes were labeled only after the fixed cells were permeabilized with 0.25% Triton X-100. These experiments showed that in the nonpermeabilizing conditions only the external GABAAR epitopes were accessible to the antibodies. Therefore, the immunolabeling of the small GABAAR clusters obtained under nonpermeabilizing conditions indicated that these GABAAR clusters (that were not associated with GABAergic synapses) were located at the cell surface and not in trafficking internal vesicles.
We have found that small GABAAR clusters and gephyrin clusters are already present at 3.5 DIV (the earliest time studied) within 2–4% of neurons. All cells that had clusters of GABAAR also had colocalized gephyrin clusters, and vice versa. At this early time point, GAD expression was very low, and therefore we could not determine whether any of the clusters were associated with GABAergic innervation. Nevertheless, immunolabeling with synaptic vesicle markers synaptophysin and SV2 suggested that GABAAR clusters were frequently localized to sites of presynaptic contacts (data not shown). This result and the observed presence of GABAAR clusters in single-cell cultures of glutamate neurons apposed to autaptic glutamate containing terminals shown by Rao et al. (2000) and confirmed in our laboratory (data not shown) indicate that small GABAAR clusters can form in the absence of any GABAergic innervation, although they are frequently associated with other presynaptic contacts.
Individual hippocampal pyramidal cells and interneurons in culture form clusters of GABAARs containing various GABAAR subunits and isoforms
Mammalian brain GABAARs are pentameric proteins composed of combinations of various subunit classes and isoforms (α1–6, β1 − 3, γ1 − 3, δ, ε, and θ) and known splice variants (i.e., γ2 long and γ2 short forms) (for review, see Barnard et al., 1998; Mehta and Ticku, 1999;Whiting et al., 1999). The most common GABAAR subunit combination found in the brain contains two α subunits, two β subunits, and one γ subunit (Im et al., 1995; Chang et al., 1996;Li and De Blas, 1997; Jechlinger et al., 1998; Farrar et al., 1999), although combinations of two α, one β, and two γ subunits also occur in the brain (Backus et al., 1993; Khan et al., 1994a,b,1996).
We have investigated the possible heterogeneity of GABAAR subunit isoform expression in both the pyramidal cells and interneurons. We observed that although most pyramidal cells and interneurons expressed the γ2 subunit (98.9%) and β2/3 subunits (99.1%), the α subunit isoforms were not expressed in all cells. Thus, α1-containing clusters were present in 69.6% of pyramidal and 73.5% of interneurons, α2clusters were present in 95.4% of pyramidal neurons and 91.0% of interneurons, and α3 was present in 53.5% of pyramidal neurons and 65.8% of interneurons. Often, more than one α subunit isoform was expressed by the same neuron (Table1).
It is worth noting that the intensity level of fluorescent signal in the receptor clusters for a particular subunit varied within and between pyramidal cells and interneurons. Thus, many of the interneurons shared very high expression levels of α1 and β2, consistent with our previous observations in the intact hippocampus (Miralles et al., 1999). The α1 subunit was also highly expressed by a subset of pyramidal neurons. The α3 subunit also showed a degree of variability in expression, with stronger signal levels seen in some interneurons and pyramidal neurons. Less variability was found in the expression of the α2, β1, β2/3, or γ2 subunits or gephyrin, because they were fully expressed by most pyramidal cells and interneurons (Table 1) (data not shown). By using single-cell PCR, others have shown that individual hippocampal neurons frequently express several α and β subunit isoform mRNAs (Brooks-Kayal et al., 1998). The aforementioned data demonstrate that this isoform heterogeneity in individual hippocampal neurons is also observed at the protein level. It is also worth mentioning that with the exception of α3, the other studied subunits are highly expressed in the intact hippocampus (Fritschy and Möhler, 1995;Sperk et al., 1997; Miralles et al., 1999; Christie et al., 2002). However, in the intact hippocampus, the α3subunit is expressed at very low levels, as shown with various antibodies, including the one used in the present study. Therefore, the cultured hippocampal neurons show upregulation of the α3 subunit expression.
The immunofluorescence labeling of the GABAAR clusters with subunit-specific antibodies represents the labeling of the complete and fully assembled GABAAR pentamers. (1) All the subunits that are necessary to form complete receptors (i.e., α, β, and γ) colocalize in the synaptic and extrasynaptic clusters (Figs. 1-3). (2) It has been shown that individual subunits or incomplete GABAARs are retained within the endoplasmic reticulum and are quickly degraded (Connolly et al., 1996; Taylor et al., 1999). (3) Only fully assembled receptor pentamers containing α and β or α, β, and γ subunits reach the cell surface (Connolly et al., 1996, 1999; Gorrie et al., 1997). In the absence of other subunits, β3 can form homopentamers that can be transported to the cell surface (Wooltorton et al., 1997; Taylor et al., 1999). However, this is unlikely to occur in the hippocampal cultures because various α isoforms and the γ2 subunit are also coexpressed by these cells. (4) The GABAAR clusters studied in the present communication are localized at the cell surface, because they are labeled by antibodies to external epitopes in nonpermeabilized cells as shown above, and (5) the data in the literature show that these cultured neurons express benzodiazepine-sensitive functional GABAARs (Segal et al., 1984; Jensen et al., 1999). The latter require the formation of pentamers and the presence of α, β, and γ2 subunits. The α and β subunits are necessary for GABA binding, whereas the α and γ2 subunits are necessary for benzodiazepine binding.
There is no segregation of GABAAR subunit isoforms to individual GABAergic synaptic clusters; however, there is partial segregation of GABAAR subunit isoforms in the small GABAAR clusters found outside GABAergic synapses
We have investigated the possibility that GABAARs with different subunit composition might be targeted to different synapses as has been reported with the α2 subunit in the intact hippocampus (Nusser et al., 1996a; Nyiri et al., 2001). For this purpose, we examined the colocalization of various GABAAR subunits and isoforms in the receptor clusters of the cells that express two isoforms. We observed complete colocalization of the α1 and α2 (Fig.2 A–C, arrows), α1 and α3 (Fig.2 D–F, arrows), β1 and β2/3 (Fig.3 A–C, arrows), and α1 and γ2 (Fig.1 G–I, arrows) subunit isoforms in all (100%) the large GABAergic synaptic clusters examined (i.e., colocalizing with GAD boutons) (Table 2). Therefore, in these cultures we have found no evidence for the segregation of receptors containing different isoforms of α or β subunits to different GABAergic synapses on the same pyramidal cell.
Regarding the small GABAAR clusters not associated with GAD, Figures 1-3 and Table 2 show that there was a high degree of colocalization of the various GABAAR subunits and isoforms with each other and with gephyrin. However, within the same cell there is also a significant proportion of small clusters that show one or the other isoform but not both (Figs.1 G,H,J,K; 2A,B,D,E; 3A,B,G,H,empty arrowheads). This result contrasts with the complete (100%) colocalization of all the studied subunits and isoforms found within the large clusters at GABAergic synapses. Thus, in neurons that express both isoforms, α1 and α2 colocalize in 58.4% of all the small clusters, whereas the remaining 41.6% of the small clusters had only α1 or α2 but not both. For α1 versus α3, 52.7% of all small clusters contained both isoforms, and the remaining 47.3% had only α1 or α3 (Table 2). These results showed that in pyramidal neurons that expressed two α subunit isoforms, 25.5–30.8% of the small clusters contained only one of the two. Therefore, in some of the small extrasynaptic GABAAR clusters, there is partial segregation of the GABAARs that contain different α subunit isoforms. This observation is consistent with the notion that the pyramidal cells express a population of GABAARs in which the two α subunits present in the pentamer are of the same isoform. Nevertheless, these cells are also very likely to have pentamers that contain two different α subunit isoforms (McKernan et al., 1991; Khan et al., 1994a,1996; Araujo et al., 1996, 1999; Jechlinger et al., 1998; Sigel and Baur, 2000). We do not know whether the small clusters that show colocalization of α1 and α2 or α1 and α3 have receptors that contain both isoforms or a mixture of receptors containing only one type of isoform. Immunocytochemistry lacks the resolution to address this issue.
The colocalization of α1 with γ2 in the small clusters was higher than that of α1 with either α2 or α3. Thus, α1 with γ2 colocalized in 68.7% of all small non-GABAergic clusters (Fig. 1 G,H, filled arrowheads), with only 14.1% of the small α1 receptor clusters not colocalized with γ2. The highest level of colocalization was observed between γ2 and gephyrin (Table 2), in which 86.0% of all small non-GABAergic clusters had both γ2 subunit and gephyrin. In this case, 94.7% of all γ2 clusters had gephyrin, and 90.0% of the gephyrin clusters had γ2 (Table 2). Some of the gephyrin clusters that did not have γ2might have γ3 (Baer et al., 1999), because pyramidal neurons also express this subunit (Wisden et al., 1992).
We also examined the aggregated colocalization of the α1, α2, and α3(α1 − 3) subunits with β2/3 and gephyrin in the small GABAAR clusters not associated with GABAergic synapses. We found that in 64.8 and 66.2% of all small clusters, α1 − 3subunits were colocalized with β2/3 and gephyrin, respectively (Table 2). We also found that 78.9% of the small clusters that contained at least one of the three α subunits also had β2/3. The α1 − 3clusters that did not colocalize with β2/3clusters might contain β1, which is also expressed by pyramidal cells of the hippocampus (see below) (Wisden et al., 1992). Table 2 also shows that 83.8% of clusters that contained at least one of the three α subunit isoforms examined colocalized with gephyrin clusters. Thus, the gephyrin clusters that did not show α1 − 3immunoreactivity might have α4 or α5 subunit-containing receptors, because pyramidal neurons also express α4 and α5 (Wisden et al., 1992). The aforementioned results support the intimate relationship between the clustering of gephyrin and the clustering of GABAARs that contain γ2 (Essrich et al., 1998; Kneussel et al., 1999). It is also worth noting that 5.3% of the small γ2 clusters and 16.2% of α1 − 3clusters did not contain gephyrin, which also supports the existence of a gephyrin-independent GABAAR clustering mechanism that operates within a small percentage of GABAAR clusters (Kneussel et al., 2001), as well as the existence of some gephyrin clusters that might not contain GABAAR clusters (Levi et al., 1999). The colocalization of the β1 or β3 subunits with the β2/3 subunits was also compared in pyramidal neurons (Fig.3 A,B,G,H). We also found that large synaptic clusters at all GABAergic synapses contained all the β subunit isoforms expressed by that neuron. These β subunit isoforms also colocalized within the majority of the small extrasynaptic clusters (Fig. 3 A,B,filled arrowheads), but not all (Fig.3 A,B,G,H,empty arrowheads).
GABAergic innervation not only induces an increase in the size of the GABAAR clusters at the GABAergic synapse, but it also leads to the disappearance of the small GABAAR clusters in the local area surrounding the GABAergic synapse
We have shown that GABAergic presynaptic inputs induce the formation of large postsynaptic GABAAR clusters at the contact sites (Figs. 1 A,D,5 A). We have also observed that this effect is accompanied by a reduction in the density of the small GABAAR clusters in dendrites that receive GABAergic innervation. We examined 50-μm-long dendrite segments, each innervated by 4–10 GAD-containing boutons, and found that there was an average of 60.6 ± 5.7% reduction (p < 0.001; n = 17) in the density of small non-GABAergic synaptic GABAAR clusters, when compared with noninnervated dendrite segments of the same neurons. Thus innervated dendrites had 27.2 ± 4.3 GABAAR clusters (mean ± SEM) in 50 μm length, whereas noninnervated dendrites had 72.2 ± 7.8 clusters in 50 μm length. We also determined that this reduction in the density of small GABAAR clusters occurs locally in the proximity of individual GABAergic synapses. Figure 5 A shows that there was a significant decrease in the density of small GABAAR clusters in the first 0–5 μm segment of dendritic length adjacent to a GABAergic synapse (1.4 ± 0.7 clusters/10 μm2; p < 0.005), as well as in the 5–10 μm segment from the synapse (1.8 ± 0.4 clusters/10 μm2;p < 0.01) and 10–15 μm segment from the synapse (3.3 ± 0.3 clusters/10 μm2;p < 0.05) compared with a control average in noninnervated dendrites of 5.5 ± 1.1 clusters/10 μm2 (Fig. 5 B). Beyond 15 μm, there was no significant difference in the density of small GABAAR clusters compared with control (15–20 μm segment: 4.7 ± 0.4/10 μm2,p = 0.23; 20–25 μm segment: 5.5 ± 0.7/10 μm2, p = 0.49).
We have also compared the density of small GABAAR and gephyrin clusters in dendrites of single pyramidal neurons in micro-island cultures with those of dendrites from pyramidal cells in mixed cultures that receive limited GABAergic innervation from interneurons. There was no significant difference in the size or density of the small GABAAR or gephyrin clusters between the dendrites of single pyramidal cells (that receive no GABAergic innervation anywhere) and the dendrites that receive no GABAergic innervation in pyramidal cells that receive GABAergic innervation in other dendrites (data not shown). These observations indicate that (1) the formation of small GABAAR clusters is not dependent on GABAergic innervation and (2) GABAergic innervation produces a reduction in the density of small clusters. Moreover, this phenomenon is a local effect that is restricted to the neighborhood of the GABAergic synapse. Therefore, the formation of large GABAAR clusters occurs exactly at the synaptic site, whereas the disappearance of the smaller clusters is a gradient effect extending to an average distance of 15–20 μm from the GABAergic synapse.
A population of small GABAAR clusters associate with glutamatergic synapses in pyramidal neurons receiving both glutamatergic and GABAergic innervation
Recently, Rao et al. (2000) reported that microcultures of isolated glutamatergic pyramidal neurons, where no GAD-containing synapses were present, showed mismatched GABAAR clusters localized postsynaptically to synaptic vesicle-containing terminals, presumably containing glutamate. We tested whether this was an anomalous situation that occurs only in the total absence of GABAergic innervation or if this also occurs in pyramidal cells that receive both GABAergic innervation from interneurons and glutamatergic innervation from themselves (autapses) and other pyramidal cells. We examined whether GABAAR clusters are associated with the postsynaptic glutamate receptor markers PSD-95 (for NMDA receptors) (Fig.6 A–C) and GluR1 (for AMPA receptors) (Fig.6 D–F). We found that 33.2% of the small non-GABAergic GABAAR clusters (19.5 ± 2.2 of 58.7 ± 7.3 total γ2 clusters/50 μm length; n = 14 matched dendrites) were juxtaposed to PSD-95 clusters (Fig. 6 A,B,filled arrowheads). Similarly, the association of GABAAR β2/3 or gephyrin clusters with AMPA receptor subunit GluR1 was also found (Fig.6 D,E, filled arrowheads). Although most of the large and small GABAAR and gephyrin clusters were localized to the dendritic shaft, the association of the small GABAAR or gephyrin clusters with GluR1 was particularly evident at the level of the dendritic spines, which were enriched in GluR1 immunoreactivity (Fig.6 D,E, arrowheads). We have also found that most (74.2 ± 9.1%) of the small GABAAR clusters that did not colocalize with GAD-positive boutons were associated with the synaptic vesicle marker SV2 (Fig. 6 G–I, filled arrowheads). Taken together, these results suggest that in pyramidal cells that receive GABAergic innervation, at least one-third of the small (non-GABAergic) GABAAR clusters are associated with glutamatergic synapses. In addition, ∼25% of small receptor clusters colocalized with neither SV2 nor GAD, suggesting that GABAAR can also form small GABAA clusters in the absence of any GABAergic or glutamatergic synaptic contact (Fig. 6 G,H,empty arrowheads).
GABAergic innervation induces the local disappearance of small GABAAR clusters, including the ones associated with glutamatergic synapses, without affecting the organization of glutamatergic receptor clusters
The aforementioned local effect of GABAergic innervation on the disappearance of small GABAAR clusters also applies to the small GABAAR clusters associated with glutamatergic synapses. After GABAergic innervation, the density of both the small GABAAR clusters associated with PSD-95 clusters and the ones not associated was decreased in the same proportion around the GABAergic synapse. This is also illustrated in Figure 6 D–F, which shows the association of GABAAR clusters with GluR1 in two dendrites of the same pyramidal cell, one with GABAergic innervation (Fig.6 D–F, empty arrowheads) and another one without (Fig. 6 D–F,filled arrowheads). In the absence of local GABAergic innervation (Fig. 6 D–F, top dendrites), the GABAAR clusters associate with the GluR1 clusters, as clearly seen on dendritic spines (6D–F, filled arrowheads). However, in the dendrite that receives GABAergic innervation (bottom dendrite), the GABAAR clusters are predominantly colocalized with the GAD-containing processes (Fig.6 D–F, arrows) instead of associating with the GluR1 clusters of dendritic spines (Fig.6 D–F, empty arrowheads).
This organizational effect was specific for GABAAR clustering because GABAergic innervation did not affect the density of PSD-95 or GluR1 clusters in the neighborhood of GABAergic synapses. There was no significant difference in the density or organization of PSD-95 clusters present in dendrites of the same pyramidal neuron that received GABAergic innervation (50.9 ± 5.0 clusters/50 μm dendrite length) with those that did not (47.9 ± 5.3 clusters/50 μm dendrite length;p = 0.34; n = 14 dendrites).
We have also found that only 4.7 + 1.4% of the PSD-95 clusters are associated with GABAergic synapses containing GAD and large synaptic GABAAR clusters. Even in this situation, there was juxtaposition of the PSD-95 clusters with GABAAR clusters at GABAergic synapses rather than true colocalization. This segregation is shown in Figure7, where the large postsynaptic GABAAR clusters follow the pattern of the GABAergic presynaptic terminal, both avoiding overlap with the postsynaptic glutamatergic receptor clusters (Fig. 7 ,insets, empty arrowheads) by taking a digitated shape rather than the normal circular or elongated form.
It has been reported previously, using similar types of hippocampal cultures, that GABAARs clusters were localized at GABA synapses but not outside GABAergic synapses (Craig et al., 1994, 1996). Moreover, most studies on GABAAR clustering do not distinguish between the two types of clusters and treat GABAAR clustering as a single phenomenon. In this communication, we report that GABAARs not only form large clusters at GABAergic synapses, but they also form small clusters outside GABAergic synapses. These small GABAAR clusters are localized at the neuronal surface, which excludes the possibility of them being intracellular trafficking receptors en route to GABAergic synaptic sites.
Our data also show that GABAergic innervation induces both the formation of large GABAAR clusters at GABAergic synapses and the disappearance of small clusters in the surrounding area. There is a gradient effect in which the disappearance of small clusters is strongest in areas immediately adjacent to GABAergic synapses and extending to 15–20 μm distance from the synapse. The greater the amount of GABAergic innervation that a pyramidal dendrite received, the higher the density of large synaptic GABAAR clusters and the lower the density of small GABAAR clusters that the dendrite showed.
Although it appears that large GABAAR clusters are formed by recruitment of smaller extrasynaptic clusters (Figs.1-7), it is unlikely that such large aggregates of receptors, gephyrin, and presumably other proteins could laterally diffuse in the membrane. Instead, we favor a mechanistic process of disassembly and assembly of the clusters. Recruitment of GABAARs to GABAergic synapses might be caused by a clustering-inducing signal generated at GABAergic synapses that induces or stabilizes the assembly of laterally mobile individual receptors into large synaptic clusters. This mechanism would also favor disassembly of small clusters present in surrounding areas by a mass action phenomenon. Meier et al. (2001) showed that glycine receptors can laterally diffuse in the membrane and be trapped into gephyrin-containing glycine receptor clusters. Nevertheless, the disappearance of small GABAAR clusters from areas surrounding GABAergic synapses could also be actively induced by a small cluster disassembly-inducing signal. For example, presynaptic terminals of the neuromuscular junction release agrin, which induces clustering of nicotinic acetylcholine receptor (nAchR) at synapses, and also provide a nerve-derived dispersal factor that disassembles extrasynaptic nAchR clusters (Lin et al., 2001). These hypothesized mechanisms do not preclude incorporation of GABAAR into clusters from internal or subsynaptic pools. Others have also reported innervation-dependent accumulation of GABAAR clusters at GABAergic synapses in cultured neurons by using immunofluorescence microscopy techniques (Craig et al., 1994; Levi et al., 1999). To the best of our knowledge, however, this is the first time that the effect of innervation on both the large synaptic GABAAR clusters and the small non-GABAergic clusters and their possible relationship has been studied.
Rao et al. (2000) reported recently that in the absence of GABAergic innervation, isolated hippocampal pyramidal cells in culture form many GABAAR clusters that were mismatched to presynaptic glutamate-containing terminals. One could argue that this is the result of an abnormal situation in which pyramidal neurons were devoid of GABAergic innervation. However, in our cultures (in which pyramidal neurons receive both GABAergic and glutamatergic innervation), a significant population of small GABAAR also associated with glutamatergic synaptic markers (GluR1, PSD-95, and synaptic vesicle marker SV2, but not GAD). This association was highest in dendrites or dendritic regions not receiving GABAergic innervation.
It is also worth noting that small GABAAR clusters associated with glutamatergic synapses were juxtaposed to GluR1 receptor clusters or PSD-95 clusters, indicating that GABAAR clusters and glutamate receptor clusters did not readily mix with each other, even if they were associated with the same glutamate-containing presynaptic terminal. These synapses contained segregated GABAARs and glutamate receptors in postsynaptic microdomains. In this respect, our results also differed from those obtained with single-cell cultures (Rao et al., 2000). They found that PSD and GluR1 receptor clusters were well separated from GABAAR clusters, proposing that either GABAARs or glutamatergic receptors (but not both) cluster in front of a single presynaptic glutamatergic terminal. Avoidance of mixing of GABAAR and glutamate receptor clusters occurred not only at glutamatergic synapses but it also occurred at GABAergic synapses (Fig.7 A–C).
We have also investigated the possible heterogeneity of GABAAR subunit composition in both large GABAAR clusters at GABAergic synapses and in small receptor clusters located outside GABAergic synapses. We found that 100% of large synaptic GABAAR clusters contained all GABAAR subunit isoforms and classes expressed by that particular neuron. However, two populations of small GABAAR clusters were found in the same neuron: 53–59% of small clusters in neurons expressing pairs of α subunit isoforms (Table 2) showed colocalization of the two isoforms, whereas the remainder of small clusters had receptors containing only one of the two isoforms. This heterogeneity in subunit isoform composition of small GABAAR clusters suggests the existence of some selectivity in clustering or trafficking of GABAAR into the small clusters.
EM immunogold studies of GABAergic synapses onto pyramidal neurons of the CA1 region of the intact hippocampus have shown a differential distribution of GABAARs containing α1 or α2 subunits in synapses made by basket cells innervating the soma and proximal dendrites of pyramidal cells (Nyiri et al., 2001). The α2 subunit concentrates postsynaptically to terminals of parvalbumin-negative basket cells and in axo-axonic synapses on the axon initial segment of hippocampal CA1 pyramidal cells (Nusser et al., 1996a). Others have also presented data at the light microscopy level consistent with differential distribution of GABAAR α1, α2, and α3 subunit isoforms in various synapses (Fritschy et al., 1992; Koulen, 1996;Fletcher et al., 1998). However, it is not entirely proven that the receptor puncta observed with the light microscope correspond to synapses. As indicated above, in our cultures, we found differential distribution of the α subunit isoforms in a proportion of the small clusters but not in large receptor clusters at GABAergic synapses (i.e., all GABAergic synapses had both α1 and α2). This is therefore different from results obtained with EM immunogold in the intact hippocampus. This difference might be attributable to selective loss of GABAergic synaptic heterogeneity in the cultures, resulting from a loss of cell types and inputs with respect to the intact brain. Some selective population of interneurons that induce and/or hinder postsynaptic accumulation of the α2-containing GABAAR in pyramidal cells might not survive the culture conditions. Interestingly, our cultures have interneurons expressing calbindin or calretinin but not parvalbumin (data not shown).
Thus we hypothesize that (1) in the intact hippocampus, GABAergic inputs from different sources have unidentified signals that are involved in targeting of postsynaptic GABAARs containing specific α subunit isoforms to specific synapses; (2) in the absence of the normal and heterogeneous GABAergic innervation and with limited GABAergic input, the various types of GABAARs containing specific subunit isoforms are pulled together to the existing GABAergic synapses; and (3) in the absence of local GABAergic innervation, GABAARs form small clusters, some of which localize at glutamatergic synapses where they remain juxtaposed to (but not displacing) glutamate receptor clusters.
It seems that GABA itself is not the signal that induces the formation of small clusters, because the latter can form in isolated glutamatergic cells, nor does activity seem to be necessary for the formation of GABAAR clusters (Craig et al., 1994;Craig, 1998), although levels of GABAAR expression are affected by neuronal activity (Penschuck et al., 1999). Formation of large GABAAR clusters at GABAergic synapses could be attributable to the presynaptic release of a molecule with a function equivalent to agrin, for nAchR clustering (Lupa and Caldwell, 1991; Lin et al., 2001) or Narp, as proposed for AMPA receptor clustering (O'Brien et al., 1999). Alternatively, direct interaction of membrane molecules from the GABAergic presynaptic terminal with membrane molecules of pyramidal neurons could trigger accumulations of large GABAergic synaptic clusters. Candidate molecules for organizing the presynaptic and postsynaptic glutamate synapse machinery at contact points are ephrins and Eph receptors (Bruckner and Klein, 1998; Torres et al., 1998), cadherins and protocadherins (Shapiro et al., 1995; Benson and Tanaka, 1998; Wu and Maniatis, 1999; Tanaka et al., 2000), and neurexins–neuroligins (Scheiffele et al., 2000). In addition, there is evidence that N-cadherins and their β-catenin partners accumulate at GABAergic synapses early in development (Benson and Tanaka, 1998). The large number of genes and alternatively spliced variants described for some of the aforementioned molecules makes them candidate molecules for also organizing the clustering of GABAARs at GABAergic synapses. Some molecules that trigger glutamate receptor clustering at glutamatergic synapses might also trigger the formation of small GABAAR postsynaptic clusters that remain juxtaposed to glutamate receptor clusters. Our results obtained with these cultures may have a bearing on processes operating in the brain, because in the cerebellum the localization of α6, γ2, and β2/3 GABAAR subunits is not restricted to GABAergic synapses. They are also present in some glutamatergic synapses (Nusser et al., 1996b, 1998). These GABAARs could participate in the regulation of glutamatergic synaptic excitability, perhaps by binding GABA that has spilled over from neighboring GABAergic synapses. Nevertheless, co-release of glutamate and GABA in some synapses has been reported (Walker et al., 2001).
This work was supported by National Institute of Neurological Disorders and Stroke Grants NS38752 and NS39287.
Correspondence should be addressed to Dr. Angel L. de Blas, 3107 Horsebarn Hill Road, U-4156, Storrs, CT 06269-4156. E-mail:.