Targeted disruption of the GABA(A) receptor delta subunit gene leads to an up-regulation of gamma 2 subunit-containing receptors in cerebellar granule cells.

GABA(A) receptors are chloride channels composed of five subunits. Cerebellar granule cells express abundantly six subunits belonging to four subunit classes. These are assembled into a number of distinct receptors, but the regulation of their relative proportions is yet unknown. Here, we studied the composition of cerebellar GABA(A) receptors after targeted disruption of the delta subunit gene. In membranes and extracts of delta-/- cerebellum, [(3)H]muscimol binding was not significantly changed, whereas [(3)H]Ro15-4513 binding was increased by 52% due to an increase in diazepam-insensitive binding. Immunocytochemical and Western blot analysis revealed no change in alpha(6) subunits but an increased expression of gamma(2) subunits in delta-/- cerebellum. Immunoaffinity chromatography of cerebellar extracts indicated there was an increased coassembly of alpha(6) and gamma(2) subunits and that 24% of all receptors in delta-/- cerebellum did not contain a gamma subunit. Because 97% of delta subunits are coassembled with alpha(6) subunits in the cerebellum of wild-type mice, these results indicated that, in delta-/- mice, alpha(6)betagamma(2) and alphabeta receptors replaced delta subunit-containing receptors. The availability of the delta subunit, thus, influences the level of expression or the extent of assembly of the gamma(2) subunit, although these two subunits do not occur in the same receptor.

previous study (18). The replacement-type DNA targeting construct placed a selectable marker gene in exon 4, upstream of the exons required for the putative transmembrane regions of this receptor subunit. This targeting event prevented the production of ␦ protein, i.e. produced a true null allele. The mice used for the present studies were the F2-F5 generation on a mixed C57BL/6J ϫ strain 129Sv/SvJ genetic background. These mice were normal in their gross behavior. ␦ϩ/ϩ mice were the F3 generation on a mixed C57BL/6J ϫ strain 129Sv/SvJ genetic background.
Quantitative Immunoblot Analysis-Membranes from a total of 12 ␦ϩ/ϩ and 12 ␦Ϫ/Ϫ cerebella were isolated individually, and equal amounts (15 g) of protein per slot were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis in the same 10% polyacrylamide gel (13). Proteins were blotted to polyvinylidene difluoride membranes and detected by subunit specific antibodies. Secondary antibodies (FabЈ 2 fragments of goat anti-rabbit IgG coupled to alkaline phosphatase, Jackson ImmunoResearch Labs Inc.) were visualized by the reaction of alkaline phosphatase with CSPD (Tropix, Bedford, MA) and the chemiluminescence signal was quantified by densitometry of Kodak X-Omat S films with the Docu Gel 2000i gel documentation system using RFLP scan software (MWG biotech, Ebersberg, Germany). The linear range of the detection system was established by determining the antibody response to a range of antigen concentrations following immunoblotting. The experimental conditions were designed such that immunoreactivities obtained in the assay were within this linear range, thus permitting a direct comparison of the amount of antigen applied per gel lane between samples. Different exposures of the same membrane were used to ensure that the measured signal was in the linear range of the x-ray film.
To test for equal protein loading, in some experiments a monoclonal anti-␤-actin antibody, was included in the antibody solution, and the amounts of endogenous ␤-actin were quantitatively determined in a way analogous to GABA A receptor subunits. Protein loading was comparable in different slots and referring the data to the amounts of endogenous ␤-actin neither changed the results nor reduced variability.
Preparation of Membrane Extracts, Affinity Chromatography, and Immunoprecipitation-GABA A receptors were solubilized from cerebellar membranes of ␦ϩ/ϩ and ␦Ϫ/Ϫ mice using a deoxycholate buffer (0.5% deoxycholate, 0.05% phosphatidylcholine, 10 mM Tris/HCl, pH 8.5, 150 mM NaCl, 1 mM benzamidine, 200 g/ml bacitracin, and 300 M phenylmethylsulfonyl fluoride), and immunoaffinity chromatography was performed as described (13). Briefly, deoxycholate extracts were cycled three times through the affinity column (synthesized as described in Ref. 21) at a rate of 2 ml/h. To determine the percentage of receptors retained by the column, immunoprecipitations with subunitspecific antibodies and [ 3 H]muscimol binding assays were performed with the original extract and the column efflux in parallel.
For immunoprecipitation, 200 l of the clear deoxycholate membrane extract was mixed with 30 l of antibody solution (20 -45 g of antibody), and the mixture was incubated under gentle shaking at 4°C overnight. Then 50 l of pansorbin (Calbiochem, La Jolla, CA) plus 100 l of a low salt buffer for immunoprecipitation (50 mM Tris/HCl, 150 mM NaCl, 1 mM EDTA, 0.2% Triton X-100, pH 8.0, 1 mM benzamidine, 200 g/ml bacitracin, and 300 M phenylmethylsulfonyl fluoride) containing 5% dry milk powder were added, and incubation was continued for 2 h at 4°C. The precipitate was centrifuged for 10 min at 10,000 ϫ g, and the pellet was washed twice with 500 l of high salt buffer for immunoprecipitation (50 mM Tris/HCl, 600 mM NaCl, 1 mM EDTA, 0.5% Triton X-100, pH 8.3, 1 mM benzamidine, 200 g/ml bacitracin, and 300 M phenylmethylsulfonyl fluoride) and once with 500 l of low salt buffer for immunoprecipitation.
For determination of the total amount of receptors present in the extract, solubilized receptors were precipitated by a mixture containing 10 g of ␤ 1 -(350 -404), 18 g of ␤ 2 -(351-405), and 15 g of ␤ 3 -(345-408) antibody per 200 l of extract. This antibody mixture was used, because all functional GABA A receptors are supposed to contain at least one of these three ␤ subunits. Precipitation with polyethylene glycol could not be used for the determination of total [ 3 H]muscimol binding due to the rapid dissociation of [ 3 H]muscimol from its binding site. The high viscosity of the polyethylene glycol solution causes relatively long fil-tration and washing times and, thus, significant losses of previously bound [ 3 H]muscimol. The amounts of ␤ 1 -(350 -404), ␤ 2 -(351-405), and ␤ 3 -(345-408) antibodies used were sufficient to maximally precipitate GABA A receptors containing the respective ␤ subunits from brain extracts. The total number of [ 3 H]muscimol binding sites measured in receptors precipitated by this antibody mixture was higher than that precipitated by polyethylene glycol, whereas the total number of [ 3 H]Ro15-4513 binding sites measured was identical whether receptors were precipitated by the antibody mixture or polyethylene glycol (13,17).
Quantification of Immunoreactivity -The animals were coded and measurements were carried out according to a double-blind protocol. Neither the person taking the images nor the person making the subsequent measurements knew the identity of the animals. The code was opened only after completing the measurements. One digital image (cooled charge-coupled device camera, Xilix Microimager, OpenLab version 2.2.1, Improvision, Coventry, UK) was captured from each of three sections per animal using a 20ϫ objective, 100-watt mercury lamp, and an L5 filter block (excitation filter, BP 480/40 nm; RKP 565 nm; suppression filter BP 610/75) of a Leitz DMRB fluorescence microscope. The same exposure time was used for images from all animals for a given antibody, and the images were treated in an identical way. The images were transferred to Adobe Photoshop (version 5.5) and the mean gray scale pixel value (12-bit) was determined separately for the molecular and granule cell layers for each section using the Histogram command. Each animal was characterized by the mean average gray scale pixel value obtained from the three sections. The values of control and ␦Ϫ/Ϫ animals were compared with the unpaired t test.
[ 3 (2). Scatchard analysis of specific [ 3 H]Ro15-4513 binding indicated that the total number (B max ) of sites was increased by 52% in ␦Ϫ/Ϫ mice, whereas the binding affinity was not significantly different (Table I). In agreement with previous reports (24 -26), 31% of total binding in the cerebellum of ␦ϩ/ϩ mice was not displaced by 100 M diazepam, but could be displaced by 100 M Ro15-1788, a benzodiazepine site antagonist. This diazepaminsensitive (DIS) binding, which is attributed to receptors containing ␣ 6 subunits (27) (Table I).
In other experiments, GABA A receptors were extracted from cerebellar membranes by a deoxycholate buffer and then immunoprecipitated with a combination of antibodies directed against the ␤ 1 , ␤ 2 , and ␤ 3 subunits of GABA A receptors (13,17).

Scatchard analysis of [ 3 H]Ro15
-4513 binding to the precipitated receptors indicated that the recovery of total, DS and DIS binding was 56%, 57%, and 53% in ␦ϩ/ϩ and 57%, 53%, and 60% in ␦Ϫ/Ϫ cerebella, respectively. Receptors recovered in the extract, however, represented 93% of all binding sites detectable in cerebellar membrane homogenates treated with detergent, because part of the receptors present in the membranes became inactivated during solubilization. Total B max values measured in cerebellar extracts from ␦Ϫ/Ϫ mice were increased by 55% as observed in the corresponding membranes. Whereas DS binding was comparable in the extracts of ␦ϩ/ϩ and ␦Ϫ/Ϫ cerebella, DIS binding in ␦Ϫ/Ϫ extracts was increased by 216% relative to that in ␦ϩ/ϩ extracts (Table I).
[ 3 H]Muscimol Binding in the Cerebellum of Control and ␦Ϫ/Ϫ Mice-To investigate the abundance of GABA A receptors, Scatchard analysis of [ 3 H]muscimol binding to cerebellar membranes of ␦ϩ/ϩ and ␦Ϫ/Ϫ mice was carried out. Comparable K d values and a slight increase in the B max value in ␦Ϫ/Ϫ mice were found, but this increase in the B max value did not reach statistical significance (Table II). Similar results were obtained from Scatchard analysis of [ 3 H]muscimol binding to receptors extracted and immunoprecipitated with a mixture of antibodies directed against the ␤ 1 , ␤ 2 , and ␤ 3 subunits, which should precipitate the vast majority of GABA A receptors (13,17): B max values measured in ␦Ϫ/Ϫ mice were higher than those found in ␦ϩ/ϩ mice, but the difference again did not reach statistical significance.
A comparison of B max values in cerebellar membranes and solubilized receptors indicated that ϳ52% and 61% of GABA A receptors present in membranes of ␦ϩ/ϩ and ␦Ϫ/Ϫ cerebella, respectively, could be recovered as binding sites in the membrane extract (Table II). The recovery of [ 3 H]muscimol binding sites was therefore comparable to that of total, diazepam-sensitive, or diazepam-insensitive [ 3 H]Ro15-4513 binding sites. The similar proportions of different binding sites recovered, as well as the parallel changes in membrane-bound and -extracted binding sites in ␦Ϫ/Ϫ mice, indicate that the receptors recovered in the extract were representative of the entire functional GABA A receptor population.

FIG. 2. Quantification of GABA A receptor subunit proteins in cerebellar membranes by Western blot analysis.
Equal amounts of cerebellar membrane proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and subunit levels were compared in ␦ϩ/ϩ and ␦Ϫ/Ϫ membranes by Western blot analysis. A, representative blot comparing immunoreactivity for ␣ 1 , ␣ 6 , ␤ 2 , ␤ 3 , and ␥ 2 subunits in three ␦ϩ/ϩ and three ␦Ϫ/Ϫ cerebella. The chemiluminescence signals were quantified by densitometry. Results are expressed as percentage of the subunit level found in ␦ϩ/ϩ membranes Ϯ S.E. B, results for ␣ 1 , ␣ 6 , ␤ 2 , ␤ 3 , and ␥ 2 subunits were obtained from 12 ␦ϩ/ϩ and 12 ␦Ϫ/Ϫ mice. Results for ␦ subunits were from three ␦ϩ/ϩ and three ␦Ϫ/Ϫ mice. No staining for the ␦ subunit could be detected in cerebellar membranes of ␦Ϫ/Ϫ mice. For statistical comparisons unpaired Student's t test was used. * p Ͻ 0.001; **p Ͻ 0.0001. of Control and ␦Ϫ/Ϫ Mice-Cerebellar membrane extracts from ␦ϩ/ϩ and ␦Ϫ/Ϫ mice were cycled on an ␣ 6 -(317-371) immunoaffinity column until ␣ 6 subunits were no longer detectable in the efflux of the column using immunoprecipitation experiments or Western blot analysis. Total GABA A receptors present in the original membrane extract and in the column efflux were then determined by [ 3 H]muscimol binding studies after precipitation with a mixture of ␤ 1 , ␤ 2 , and ␤ 3 subunitspecific antibodies. The total amount of GABA A receptors in the column efflux was reduced by 55.7% in ␦ϩ/ϩ and by 57.5% in ␦Ϫ/Ϫ mice, indicating that the proportion of ␣ 6 subunit-containing receptors present in the extracts from these tissues was similar (Table III).
In other experiments, the extent of reduction of ␣ 1 , ␤ 2 , ␤ 3 , ␥ 2 , or ␦ subunit-containing receptors after chromatography on the anti-␣ 6 column was determined. For this, receptors containing these subunits were immunoprecipitated in the original extract and the anti-␣ 6 column efflux with ␣ 1 , ␤ 2 , ␤ 3 , ␥ 2 , or ␦ subunitspecific antibodies and were quantitatively determined by [ 3 H]muscimol binding. In the absence of any cross-reactivity of the anti-␣ 6 antibodies with other GABA A receptor subunits (13,17), the percentage of reduction indicates the percentage of these receptors containing ␣ 6 subunits. As shown in Table III, the percentage of ␣ 1 , ␤ 2 , ␤ 3 , ␥ 2 , or ␦ subunit-containing receptors containing ␣ 6 subunits in the cerebellum of ␦ϩ/ϩ mice was similar to that reported previously (17) for the cerebellum of wild-type mice.
The percentage of ␣ 1 , ␤ 2 , and ␤ 3 subunit-containing receptors removed by the anti-␣ 6 column from extracts of ␦Ϫ/Ϫ cerebella was comparable, indicating that a similar proportion of these receptors contained ␣ 6 subunits in control and ␦Ϫ/Ϫ cerebella (Table III).
In contrast, the percentage of ␥ 2 receptors eliminated by the anti-␣ 6 column was significantly larger in the cerebellum of ␦Ϫ/Ϫ mice, suggesting an increased formation of GABA A receptors containing both ␣ 6 and ␥ 2 subunits. The actual increase in the number of receptors containing these subunits, however, was even larger, because of the increased expression of ␥ 2 subunits (Fig. 2) and the increased number of ␥ 2 subunitcontaining receptors in ␦Ϫ/Ϫ cerebellum (Table I).
To further investigate the composition of ␥ 2 -containing receptors, these receptors were completely eliminated from cerebellar extracts of ␦ϩ/ϩ and ␦Ϫ/Ϫ mice using an immunoaffinity column containing the ␥ 2 -(319 -366) antibody. This column eliminated 68.3% of all receptors in ␦ϩ/ϩ and 76.5% of all receptors in ␦Ϫ/Ϫ mice. These proportions were not signifi-cantly different (Table IV). It has to be kept in mind, however, that when comparing the number of ␥ 2 -containing receptors present in ␦ϩ/ϩ and ␦Ϫ/Ϫ mice, the total number of GABA A receptors in these tissues has to be considered (Table II).
Quantification in the original membrane extract and column efflux of receptor subtypes precipitated with subunit-specific antibodies indicated that a comparable proportion of ␣ 1 subunit-containing receptors was removed by the anti-␥ 2 affinity column in ␦ϩ/ϩ and ␦Ϫ/Ϫ extracts. In agreement with the results from the anti-␣ 6 column (Table III) a significantly higher percentage of ␣ 6 receptors was associated with ␥ 2 subunits in extracts from ␦Ϫ/Ϫ cerebella (Table IV). Although only 52.4% of ␣ 6 receptors were eliminated by the anti-␥ 2 subunit column from membrane extracts of ␦ϩ/ϩ mice, 70.2% of ␣ 6 receptors were eliminated from those of ␦Ϫ/Ϫ mice. The extent of association with ␥ 2 subunits was also slightly elevated for ␤ 2 subunits: 67% of ␤ 2 subunit-containing receptors contained ␥ 2 subunits in ␦ϩ/ϩ and 75.2% in ␦Ϫ/Ϫ cerebella. The degree of association of ␤ 3 and ␥ 2 subunits, however, was comparable in the cerebellum of ␦ϩ/ϩ and ␦Ϫ/Ϫ mice (Table IV).
Supporting previous conclusions (13,28,29), the percentage of ␦ subunit-containing receptors present in cerebellar extracts from ␦ϩ/ϩ mice (28.8 Ϯ 4.2% of all receptors) was not reduced in the efflux of the anti-␥ 2 subunit column (Table IV). This demonstrates that GABA A receptors containing both ␥ 2 and ␦ subunits are not present in the cerebellum.
Summing of the proportions of ␦ (28.8%)-and ␥ 2 (68.3%)containing receptors in ␦ϩ/ϩ mice indicates that 97.1% of all receptors in the cerebellum are assemblies of ␣␤␥ 2 or ␣␤␦ combinations. Interestingly, despite the absence of ␦ subunits, only 76.5% of all receptors in ␦Ϫ/Ϫ mice contained ␥ 2 subunits. This raised the possibility that the remaining GABA A receptors contained up-regulated ␥ 1 or ␥ 3 subunits. However, no significant amounts of ␥ 1 or ␥ 3 subunit-containing GABA A receptors could be immunoprecipitated in cerebellar extracts from ␦ϩ/ϩ and ␦Ϫ/Ϫ mice using ␥ 1 -(324 -366) or ␥ 3 -(322-372) antibodies under conditions where these antibodies precipitated significant amounts of receptors in forebrain extracts (experiments not shown, Ref. 21). This indicated that 23.5% of all GABA A receptors in cerebellum extracts from ␦Ϫ/Ϫ mice did not contain ␥ 1 , ␥ 2 , ␥ 3 , or ␦ subunits. Because significant amounts of ␣ 1 , ␣ 6 , ␤ 2 , or ␤ 3 subunit-containing receptors remained in the extract after complete elimination of ␥ 2 subunit-containing  from ␦ϩ/ϩ and ␦Ϫ/Ϫ mice GABA A receptors were extracted from the cerebellum of ␦ϩ/ϩ and ␦Ϫ/Ϫ mice, and ␣ 6 subunit-containing receptors were completely eliminated from the extract by chromatography on an immunoaffinity column containing the antibody ␣6-(317-371). GABA A receptors containing ␣ 1 , ␤ 2 , ␤ 3 , ␥ 2 , or ␦ subunits were precipitated in the original extract and in the column efflux using the respective subunit-specific antibodies. Precipitated receptor subtypes were quantified using [ 3 H]muscimol binding. The percentage of reduction in the column efflux of receptors containing the respective subunit indicated their extent of colocalization with ␣ 6 subunits in the same receptors. Data are mean values Ϯ S.E. with the numbers of experiments given in parentheses. Results from ␦Ϫ/Ϫ mice were statistically compared with results from ␦ϩ/ϩ mice using Student's unpaired t test. receptors (Table IV), it can be concluded that the remaining 23.5% of GABA A receptors were composed of combinations of ␣ 1 , ␣ 6 , ␤ 2 , and ␤ 3 subunits, or additionally contained as yet unidentified subunits.

No Reduction in the Number of GABA A Receptors in ␦Ϫ/Ϫ
Cerebellum-In the present study possible changes in the abundance and subunit composition of GABA A receptors were investigated in the brain of mice with a targeted disruption of the ␦ subunit expression. Because the ␦ subunit is much more abundant in the cerebellum than in the forebrain (30, 31), we focused our investigation on this brain region. Despite the loss of ␦ subunit-containing receptors, which comprise nearly 30% of all GABA A receptors in the cerebellum, the number of GABA A receptors was not reduced in membranes or membrane extracts of ␦Ϫ/Ϫ cerebellum, as demonstrated by Scatchard analysis of [ 3 H]muscimol binding. These data are in contrast to those of a previous study (18) in which a significant reduction in [ 3 H]muscimol binding was observed in whole brain homogenates of ␦Ϫ/Ϫ mice. The different tissues used, as well as differences in the tissue preparation and [ 3 H]muscimol binding assay conditions, might have contributed to the different results.
Increase in GABA A Receptors Containing ␥ 2 Subunits in ␦Ϫ/Ϫ Cerebellum-Western blot analysis demonstrated a 45% increase in the expression of ␥ 2 subunits in the cerebellum of ␦Ϫ/Ϫ mice, which corresponded to a 52% increase in GABA A receptors containing [ 3 H]Ro15-4513 binding sites. This finding is consistent with the elevated [ 3 H]Ro15-4513 binding in cerebellum and several other brain tissues of ␦Ϫ/Ϫ mice observed in autoradiographic studies (18). The increase in [ 3 H]Ro15-4513 binding sites was due to a dramatic increase in DIS binding sites in membranes and extracts from ␦Ϫ/Ϫ cerebellum. Because DIS binding sites are formed by receptors containing ␣ 6 ␤␥ 2 or ␣ 1 ␣ 6 ␤␥ 2 subunits (12,25,27), these data suggested a significant increase in the formation of receptors containing ␣ 6 as well as ␥ 2 subunits in the cerebellum of ␦Ϫ/Ϫ mice. This conclusion was supported by studies investigating subunit partnerships in GABA A receptors: Immunoaffinity chromatography using an anti-␣ 6 subunit column showed that the proportion of ␥ 2 receptors containing ␣ 6 subunits was significantly larger in membrane extracts from ␦Ϫ/Ϫ than in those from ␦ϩ/ϩ mice. Similarly, the use of an anti-␥ 2 subunit column showed the increased coassembly of the ␥ 2 and ␣ 6 subunits but not the ␥ 2 and ␣ 1 subunits in extracts from ␦Ϫ/Ϫ cerebella. These data indicated an increased formation of ␣ 6 ␤␥ 2 but not of ␣ 1 ␣ 6 ␤␥ 2 receptors in ␦Ϫ/Ϫ cerebellum.
Increase in GABA A Receptors Composed of ␣␤ Subunits in ␦Ϫ/Ϫ Cerebellum-In wild-type cerebellum, 68.3% and 28.8% of GABA A receptors contained ␥ 2 and ␦ subunits, respectively. Because ␥ 2 and ␦ subunits are not colocalized in the same GABA A receptors in this tissue (13,28) 97% of all cerebellar GABA A receptors contained ␥ 2 or ␦ subunits. This is consistent with the finding that receptors containing ␥ 1 or ␥ 3 subunits are not very abundant in the cerebellum (30,31). After the complete loss of ␦ receptors, most of the remaining receptors might be expected to contain ␥ 2 subunits. However, only 76.5% of all receptors in the extract of ␦Ϫ/Ϫ cerebella contained ␥ 2 subunits. Because ␥ 1 or ␥ 3 receptors were not up-regulated in the cerebellum of ␦Ϫ/Ϫ mice, the remaining receptors must have been composed of either ␣ and ␤ subunits only, or these subunits may have been associated with as yet unidentified subunits. Because comparable proportions of ␣ 1 , ␣ 6 , ␤ 2 , or ␤ 3 subunit-containing receptors remained in the extracts of ␦Ϫ/Ϫ cerebellum after the removal of ␥ 2 subunit-containing receptors, the remaining receptors were probably composed of various combinations of ␣ 1 , ␣ 6 , ␤ 2 , and ␤ 3 subunits. The exact composition of receptors formed, however, cannot be deduced from the data available.
Changes in the Subunit Composition of GABA A Receptors in ␦Ϫ/Ϫ Cerebella Mostly Occur in Granule Cells-Immunohistochemical observations indicated a large increase in labeling for the ␥ 2 subunit in the cerebellar granule cell layer and a small increase in the molecular layer of ␦Ϫ/Ϫ mice. Combined with the fact that in cerebellum ␣ 6 subunits are exclusively expressed in granule cells (where they are colocalized with ␦ subunits in wild-type mice) and with the observation suggesting an increased formation of ␣ 6 ␤␥ 2 receptors in ␦Ϫ/Ϫ mice, this indicates that most of the changes in receptor composition observed in mice with a targeted disruption of the ␦ subunit were induced in granule cells.
Although a significant increase in the expression of ␣ 1 and ␤ 3 subunits was observed in Western blot studies, no apparent change in the expression of these subunits was observed in immunocytochemical studies in ␦Ϫ/Ϫ cerebella. Because changes in ␣ 1 and ␤ 3 subunit expression were small and varied in different animals, they might have escaped detection in immunocytochemical studies. Because there was no change in the number of DS [ 3 H]Ro15-4513 binding sites in the cerebellum of ␦Ϫ/Ϫ mice, and because the extent of coassembly of ␣ 1 and ␤ 3 subunits with ␥ 2 receptors was comparable in extracts from ␦ϩ/ϩ and ␦Ϫ/Ϫ cerebella, the increased abundance of ␣ 1 and ␤ 3 subunits did not result in an increased formation of ␣ 1 ␤ 3 ␥ 2 receptors. These subunits, therefore, either formed ␣ 1 ␤ 3 receptors that contributed to the number of [ 3 H]muscimol but not of [ 3 H]Ro15-4513 binding sites (32) or represented unassembled ␣ 1 and ␤ 3 subunits. The first possibility seems to be more likely, because specific [ 3 H]muscimol binding in cerebellar membranes or membrane extracts was slightly, although not significantly, increased in ␦Ϫ/Ϫ mice.
Molecular Mechanism Causing the Changes in GABA A Receptor Subunit Composition in ␦Ϫ/Ϫ Cerebellum-Immunoaffinity purification on an anti-␣ 6 subunit column showed that ␦ subunits almost exclusively are coassembled with ␣ 6 subunits. The corresponding GABA A receptors are composed of ␣ 6 ␤␦ and ␣ 1 ␣ 6 ␤␦ subunits in the rat (13). Because there was no change in the expression of ␣ 6 subunits in the cerebellum of ␦Ϫ/Ϫ mice, ␣ 6 ␤␦ and ␣ 1 ␣ 6 ␤␦ receptors presumably were converted into from ␦ϩ/ϩ and ␦Ϫ/Ϫ mice GABA A receptors were extracted from the cerebellum of ␦ϩ/ϩ and ␦Ϫ/Ϫ mice, and ␥ 2 subunit-containing receptors were completely eliminated from the extract by chromatography on an immunoaffinity column containing the antibody ␥ 2 -(319 -366). GABA A receptors containing ␣ 1 , ␣ 6 , ␤ 2 , ␤ 3 , or ␦ subunits were precipitated in the original extract and in the column efflux using the respective subunit-specific antibodies. Precipitated receptor subtypes were quantified using [ 3 H]muscimol binding. The percentage of reduction in the column efflux of receptors containing the respective subunit indicated their extent of colocalization with ␥ 2 subunits in the same receptor. Data are mean values Ϯ S.E. with the numbers of experiments given in parentheses. Results from ␦Ϫ/Ϫ mice were statistically compared with results from ␦ϩ/ϩ mice using Student's unpaired t test. ␣ 6 ␤, ␣ 1 ␣ 6 ␤, and ␣ 6 ␤␥ 2 receptors. Studies on GABA A receptor assembly in non-neuronal cells have indicated that ␣ and ␤ subunits form subunit tetramers and pentamers, whereas ␣, ␤, and ␥ subunits exclusively form pentamers (20). The rate of formation of ␣␤ pentamers, therefore, presumably is significantly slower than that of ␣␤␥ (and possibly also of ␣␤␦) pentamers (33,34). Once formed, however, ␣␤ pentamers are transported to the cell surface (34) and form functional receptors (35). It is, therefore, possible that most of the ␣␤ subunit combinations that do not assemble with ␥ 2 subunits in the cerebellum of ␦Ϫ/Ϫ mice represented completely assembled and functional receptors. However, the presence of functional ␣␤ receptor on the surface of these cells needs to be directly demonstrated.
Comparison with Previous Studies on Targeted Disruption of GABA A Receptor Subunits-Previous studies have indicated that the targeted disruption of ␥ 2 , ␤ 3 , or ␣ 6 subunits did not induce a compensatory up-regulation of the transcription of other GABA A -receptor subunit genes in mutant mouse brain (16, 36 -38). But these studies also indicated that the availability of ␣ and ␤ subunits was essential for receptor formation. Thus, upon disruption of the ␣ 6 or ␤ 3 subunit expression, the extent of reduction in the number of the remaining receptors indicated a complete loss of receptors containing the respective subunit and a corresponding reduction in the number of the remaining GABA A receptors (39).
Upon disruption of the ␥ 2 subunit expression, however, neither the number of GABA A receptors nor the protein level of the subunits ␣ 1 , ␣ 2 , ␣ 3 , ␤ 2/3 , were reduced, indicating the predominant formation of receptors composed of ␣␤ subunits (36). In addition, the protein level of ␥ 1 and ␥ 3 subunits remained unaltered in ␥ 2 Ϫ/Ϫ mice as shown by Western blotting. This may indicate that the supply of ␥ 1 and ␥ 3 subunits is not linked to the availability of ␣␤ subunits, because receptors containing these subunits either are not synthesized in the same cells or the same subcellular compartments as those containing ␥ 2 subunits. The extent of expression of ␦ subunits was not investigated in this study.
The likely presence of receptors composed of ␣␤ subunits without additional subunit classes in the cerebellum of ␦Ϫ/Ϫ mice is consistent with the expression of these receptors in ␥ 2 Ϫ/Ϫ mice (36). The additional up-regulation of ␣ 6 ␤␥ 2 receptors in ␦Ϫ/Ϫ mice can be explained if the level of ␥ 2 subunit expression is linked to the function of the granule cells. The latter mechanism might be unique for ␥ 2 subunits, which are the most abundant GABA A receptor subunits in the brain and seem to be responsible for receptor anchoring (40,41). Alternatively, it is possible that in wild-type mice ␥ 2 and ␦ subunits compete for ␣ 6 and ␤ subunits during assembly of receptors and that subunits unable to find assembly partners are degraded (34). A lack of ␦ subunits in ␦Ϫ/Ϫ mice might then reduce degradation of the surplus ␥ 2 subunits by increasing their chances to find assembly partners, thus leading to an increased formation of ␣ 6 ␤␥ 2 receptors. Future studies will have to decide which of these possibilities most likely occur.