Detection and binding properties of GABA(A) receptor assembly intermediates.

Density gradient centrifugation of native and recombinant gamma-aminobutyric acid, type A (GABA(A)) receptors was used to detect assembly intermediates. No such intermediates could be identified in extracts from adult rat brain or from human embryonic kidney (HEK) 293 cells transfected with alpha(1), beta(3), and gamma(2) subunits and cultured at 37 degrees C. However, subunit dimers, trimers, tetramers, and pentamers were found in extracts from the brain of 8-10-day-old rats and from alpha(1)beta(3)gamma(2) transfected HEK cells cultured at 25 degrees C. In both systems, alpha(1), beta(3), and gamma(2) subunits could be identified in subunit dimers, indicating that different subunit dimers are formed during GABA(A) receptor assembly. Co-transfection of HEK cells with various combinations of full-length and C-terminally truncated alpha(1) and beta(3) or alpha(1) and gamma(2) subunits and co-immunoprecipitation with subunit-specific antibodies indicated that even subunits containing no transmembrane domain can assemble with each other. Whereas alpha(1)gamma(2), alpha(1)Ngamma(2), alpha(1)gamma(2)N, and alpha(1)Ngamma(2)N, combinations exhibited specific [(3)H]Ro 15-1788 binding, specific [(3)H]muscimol binding could only be found in alpha(1)beta(3) and alpha(1)beta(3)N, but not in alpha(1)Nbeta(3) or alpha(1)Nbeta(3)N combinations. This seems to indicate that a full-length alpha(1) subunit is necessary for the formation of the muscimol-binding site and for the transduction of agonist binding into channel gating.

whereas misfolded or improperly oligomerized subunits are retained in the endoplasmatic reticulum and degraded (2,3,5). Little is known about the molecular events involved in subunit oligomerization and formation of ligand-binding sites. In the present study the first steps of these events are investigated for GABA A receptors.
GABA A receptors are chloride channels that can be opened by GABA (6) and are the site of action of various pharmacologically and clinically important drugs, such as benzodiazepines, barbiturates, steroids, anesthetics, and convulsants. These drugs modulate GABA-induced chloride flux by interacting with separate and distinct allosteric binding sites (7). So far, at least 19 GABA A receptor subunits belonging to several subunit classes (six ␣, three ␤, three ␥, one ␦, one ⑀, one , one , and three ) have been identified in the mammalian brain (8,9). Expression studies indicated that ␣, ␤, and ␥ subunits have to combine to form GABA A receptors with a pharmacology resembling that of the majority of native receptors (7). Most reports agree that these receptors are composed of two ␣, two ␤, and one ␥ subunit (10 -13).
Density gradient centrifugation studies indicated that recombinant GABA A receptors composed of ␣ 1 ␤ 3 ␥ 2 subunits almost exclusively sediment as subunit pentamers. ␣ 1 ␤ 3 subunit combinations sediment as tetramers and pentamers, whereas combinations of ␣ 1 ␥ 2 or ␤ 3 ␥ 2 subunits predominantly form heterodimers (12). These results suggested a subunit arrangement in GABA A receptors in which four alternating ␣ and ␤ subunits are connected by a ␥ subunit (12).
Presently, however, nothing is known about the processes that lead from single subunits to completely assembled and pharmacologically functional receptors. Because no assembly intermediates could be identified in HEK cells transfected with ␣ 1 ␤ 3 ␥ 2 subunits under the conditions used and because not all of the subunit dimers that can be formed in HEK cells transfected with two different subunits might be formed when all three subunits are co-expressed, it is not clear whether ␣␤, ␣␥, or ␤␥ subunit dimers or some or all of these dimers are the starting point for GABA A receptor synthesis.
The pentameric receptor possesses binding sites for the endogenous neurotransmitter GABA, presumably located at the interface between ␣ 1 and ␤ 3 subunits (14), for benzodiazepines, located between the ␣ 1 and ␥ 2 subunit (15), as well as for TPBS, presumably located either within or close to the channel formed by these subunits (16 -18). Presently, nothing is known about the events leading to the formation of the various binding sites on GABA A receptors.
By lowering the temperature during culture of HEK cells transfected with ␣ 1 , ␤ 3 , and ␥ 2 subunits, in the present study we were able to detect assembly intermediates of GABA A receptors using sucrose density gradient centrifugation. Results indicated that different subunit dimers are formed during GABA A receptor assembly. Studies investigating the dimerization of complete and truncated subunits additionally suggested that the binding sites for [ 3
Generation of cDNA Constructs-For the generation of recombinant receptors, ␣ 1 , ␤ 3 , and ␥ 2 subunits of GABA A receptors from rat brain were cloned and subcloned into pCDM8 expression vectors (Invitrogen, San Diego, CA) as described previously (12,23).
Truncated subunits were constructed by polymerase chain reaction amplification using the full-length subunit as template. The polymerase chain reaction primers contained EcoRI and HindIII restriction sites, which were used to clone the fragments into pCDNAIAmp vectors (Invitrogen). The truncated subunits were confirmed by sequencing.
Density Gradient Centrifugation-Transfected HEK cells were incubated 44 h at 37°C or 8 h at 37°C followed by 16 h at 25°C. Cells from eight culture dishes were harvested and extracted in 1.6 ml of Lubrol extraction buffer (1% Lubrol PX, 0.18% phosphatidylcholine, 150 mM NaCl, 5 mM EDTA, 50 mM Tris-HCl, pH 7.4, containing 0.3 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, and 100 mg/liter bacitracin) for 8 h at 4°C. This buffer was used rather than a Triton X-100 or a deoxycholate buffer. Because of its low solubilizing ability it did not dissociate assembly intermediates and, thus, allowed their identification (3). Membranes from adult rat brains (preparation described in Ref. 25) were extracted in 3.5 ml of Lubrol extraction buffer/brain. The extracts were centrifuged for 40 min at 150,000 ϫ g at 4°C, and 200 l of the extracts was layered onto the top of a density gradient 5-20% sucrose in Lubrol extraction buffer). For the determination of sedimentation coefficients, 2 g of digoxygenated catalase (sedimentation coefficient, 11 s), 1.2 g of digoxygenated alkaline phosphatase (sedimentation coefficient, 6.1 s), and 1 g of digoxygenated carbonic anhydrase (sedimentation coefficient 3.3 s) were included in the overlays. The gradients were centrifuged at 120,000 ϫ g at 4°C for 23 h and were then fractionated by piercing at the tube bottom (12). Protein in individual fractions was precipitated (26) and dissolved in sample buffer (108 mM Tris-sulfate, pH 8.2, 10 mM EDTA, 25% (w/v) glycerol, 2% SDS, and 3% dithiothreitol) for SDS-PAGE. Proteins were identified by Western blot analysis, and signal intensity per fraction was quantified as described below. In a previous study (12) it has been demonstrated that after co-transfection of HEK cells with ␣ 1 , ␤ 3 , and ␥ 2 subunits all three subunits sedimented at a single peak of 8.7 s, representing the pentameric receptor. After co-transfection of HEK cells with ␣ 1 and ␤ 3 subunits, both subunits again sedimented at a peak of 8.7 s. The ␤ 3 subunit, however, additionally formed subunit complexes that sedimented at 3.3, 5.5, 6.7, and 7.4 s (12). Assembly intermediates with similar sedimentation properties have been observed previously for the homologous nAChR using a similar procedure (3). Thus, the monomeric subunits of these receptors exhibited a sedimentation between 3 and 4 s, sedimentation of subunit dimers was observed at 6 s, trimers sedimented at 7 s, tetramers sedimented at 8 s, and pentamers sedimented at 9 s (3,12). From this it was concluded that the 3.3-, 5.5-, 6.7-, and 7.4-s peaks of GABA A receptor assembly intermediates represent subunit mono-, di-, tri-, and tetramers (3).
Purification and Co-immunoprecipitation of GABA A Receptor Subunits-Transfected HEK cells were incubated 44 h at 37°C. Cells from four culture dishes were extracted with 800 l of Lubrol extraction buffer for 8 h at 4°C. The extract was centrifuged for 40 min at 150,000 ϫ g at 4°C, and the clear supernatant was incubated overnight at 4°C under gentle shaking with 15 g ␤ 3 (345-408) or ␥ 2 (319 -366) antibodies. After addition of Immunoprecipitin (preparation described in Ref. 12 and 0.5% nonfat dry milk powder and shaking for additional 3 h at 4°C, the precipitate was washed three times with IP low buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, pH 8.0) containing 1% Triton X-100. The precipitated proteins were dissolved in sample buffer and subjected to SDS-PAGE and Western blot analysis.
Radioligand Binding Studies-For binding studies frozen membranes from untransfected or transfected HEK cells were thawed, and cells were homogenized in 50 mM Tris/citrate buffer, pH 7.4, by using an Ultraturax, followed by three centrifugation (200 000 ϫ g for 20 min at 4°C) resuspension cycles. Cell pellets were resuspended in 50 mM Tris/citrate buffer, pH 7.4, at a protein concentration in the range of 0.5-1 mg/ml as measured with the BCA protein assay kit (Pierce) with bovine serum albumin as standard. Membranes were then incubated for 90 min at 4°C in a total of 1 ml of a solution containing 50 mM Tris/citrate buffer, pH 7. Membranes were then filtered through Whatman GF/B filters, and the filters were rinsed twice with 3.5 ml of ice-cold 50 mM Tris/citrate buffer and were then subjected to scintillation counting.  (23).
Immunofluorescence-HEK cells were fixed with 2% paraformaldehyde in PBS 30 -35 h after transfection, followed by a 10-min wash in 50 mM NH 4 Cl in PBS. Washes between incubation steps were performed in PBS. For detection of intracellular receptors, cells were permeabilized with 0.1% Triton X-100 for 5 min. Blocking was performed in 5% bovine serum albumin in PBS for 10 min, followed by an incubation with primary antibody in 1% bovine serum albumin in PBS. Primary antibodies were detected with goat anti-rabbit IgG (HϩL) Bodipy FL (Molecular Probes, Eugene, OR) or donkey anti-mouse IgG (HϩL) Cy3 (Amersham Pharmacia Biotech) in 1% bovine serum albumin in PBS. Labeling was visualized using a Zeiss Axiovert 135 M microscope attached to a confocal laser system (Carl Zeiss LSM 410), equipped with an argon laser and a helium-neon laser and suitable filter sets. To verify that labeling of cells without permeabilization was restricted to the cell surface, parallel samples were stained with antibodies directed against the intracellular loop of GABA A receptor subunits (experiments not shown). These antibodies detected GABA A receptor subunits only after permeabilization of transfected cells. Results obtained from double labeling experiments were compared with single labeling experiments to demonstrate that the labeling pattern in double labeling experiments was not caused by cross-bleeding artifacts (experiments not shown).

Detection of Assembly
Intermediates-In an attempt to identify GABA A receptor assembly intermediates, extracts from adult rat brain or from HEK 293 cells transfected with ␣ 1 , ␤ 3 , and ␥ 2 subunits were subjected to sucrose density gradient centrifugation. Under these conditions, depending on their molecular mass, monomeric and multimeric proteins migrate into the gradient with different sedimentation coefficients. Gradients were fractionated, and the proteins in individual fractions were precipitated and subjected to SDS-PAGE and Western blot analysis with subunit specific antibodies. s values of receptors and receptor intermediates were determined by analyzing the sedimentation of standard proteins with known s values added to each gradient.
As shown in Fig. 1A for brain extracts and Fig. 1B for transfected HEK cells, the ␣ 1 , ␤ 3 , as well as the ␥ 2 subunit proteins sedimented at a single peak at 8.7 s. This sedimentation coefficient has been reported to represent the completely assembled pentameric GABA A receptor, and the protein shoulder above 8.7 s presumably is caused by an aggregation of pentameric receptors (3,12). The absence of ␣ 1 , ␤ 3 , and ␥ 2 protein peaks with lower s values ( Fig. 1) indicated that most of the GABA A receptors formed in adult brain as well as in transfected cells are pentamers and that receptor synthesis in these tissues was low and/or assembly of receptors was too fast to allow an identification of assembly intermediates under these conditions.
In other experiments, the culture temperature was first kept at 37°C for the first 8 h after transfection of HEK cells with ␣ 1 , ␤ 3 , and ␥ 2 subunits to normally initiate transcription and translation and was then reduced to 25°C for the following 16 h to slow down the assembly of recombinant GABA A receptors. Receptors formed were extracted from the cells and were subjected to density gradient centrifugation. As shown in Fig.  2, under these conditions assembly intermediates could be identified. ␣ 1 and ␥ 2 subunits sedimented at 3.3, 4.5, 5.5, 6.8, 7.4, and 8.7 s (Fig. 2). The sedimentation pattern of the ␤ 3 subunit was similar to that of the ␣ 1 and ␥ 2 subunits, but the protein peak at 6.8 s could not be identified and presumably was present in the shoulder of the 7.4-s peak. It has been reported previously (12) that the peaks at 3.3, 5.5, 6.7, 7.4, and 8.7 s represent mono-, di-, tri-, tetra-, and pentamers of GABA A receptor subunits, respectively. The additional peak at 4.5 s could represent a monomer bound to a chaperone, because recently it has been suggested that chaperones might stabilize subunit monomers (4). The identification of all these protein peaks by all three antibodies was not due to a cross-reactivity FIG. 1. Sucrose density gradient centrifugation of GABA A receptors. Receptors from adult rat brain (A) or from HEK cells transfected with ␣ 1 , ␤ 3 , and ␥ 2 subunits (B) were extracted and centrifuged on 5-20% linear sucrose density gradients. Gradients were fractionated, and proteins in individual fractions were precipitated and subjected to SDS-PAGE and Western blot analysis using ␣ 1 (1-9), ␤ 3 (345-408), and ␥ 2 (319 -366) antibodies. s values were measured by including digoxigenized standard proteins with known s values in each gradient. OD, optical density (arbitrary units); s, sedimentation value. The experiments were performed four times with comparable results.
FIG. 2. Sucrose density gradient centrifugation of GABA A receptors after culturing of transfected HEK cells at 25°C. HEK cells were transfected with ␣ 1 , ␤ 3 , and ␥ 2 subunits and were incubated for 8 h at 37°C and afterward for 16 h at 25°C. Extracts of these cells were subjected to sucrose density gradient centrifugation as described in the legend to Fig. 1. Individual fractions of the gradients were analyzed in Western blots using ␣ 1 (1-9) (A), ␤ 3 (345-408) (B), and ␥ 2 (319 -366) (C) antibodies. OD, optical density (arbitrary units); s, sedimentation value. The experiments were performed eight times with comparable results. of the antibodies, because none of the antibodies used for these experiments exhibited any cross-reactivity with other subunits as demonstrated by Western blot analysis of various recombinant receptors (27,28).
In other experiments, the sedimentation properties of GABA A receptors extracted from the brain of 8 -10-day-old rats were investigated. In this developing tissue, GABA A receptors are continuously synthesized in different neurons, and it was hoped that amounts of assembly intermediates sufficient to be detected would be present. As shown in Fig. 3, this actually was the case: ␣ 1 , ␤ 3 , and ␥ 2 subunits extracted from 8 -10-day-old rats, in contrast to those from adult rat brain, sedimented in multiple, overlapping peaks. Whereas the sedimentation pattern of ␣ 1 and ␥ 2 subunits was again similar, showing overlapping peaks and shoulders at 5.5, 6.8, and 8.7 s, the sedimentation pattern of ␤ 3 subunits was slightly different, showing prominent peaks at 3.3 and 5.5 s and overlapping peaks and shoulders between 6.8 and 8.7 s (Fig. 3).

Formation of the [ 3 H]Muscimol-binding
Site-Because the [ 3 H]muscimol-binding site on GABA A receptors is located at the interface of ␣ and ␤ subunits (14), it was interesting to investigate whether this binding site could already be formed by GABA A receptor assembly intermediates containing ␣ 1 and ␤ 3 subunits. Because co-transfection of HEK cells with ␣ 1 and ␤ 3 subunits leads to the formation of ␣ 1 ␤ 3 tetramers and pentamers (12), truncated ␣ 1 (␣ 1 N) and ␤ 3 (␤ 3 N) subunits contain-ing the complete extracellular N-terminal domain but no transmembrane domains were cloned to investigate whether they can assemble with full-length ␤ 3 and ␣ 1 subunits, respectively, forming smaller assembly intermediates. ␣ 1 and ␤ 3 subunits and ␣ 1 N and ␤ 3 N fragments were then co-transfected into HEK cells in various combinations, and expressed subunits were extracted from these cells and were immunoprecipitated with ␤ 3 (1-13) antibodies. The precipitate was subjected to SDS-PAGE and Western blot analysis using digoxigenized ␣ 1 (1-9) antibodies. As shown in Fig. 4A, in extracts from HEK cells co-transfected with full-length ␣ 1 and ␤ 3 subunits or full-length ␣ 1 and ␤ 3 N subunits, a strongly labeled protein band with apparent molecular mass 51 kDa was detected in Western blots. A protein band with identical molecular mass could be precipitated by ␣ 1 (1-9) antibodies from these cells as well as from HEK cells transfected with ␣ 1 subunits only, but not from untransfected HEK cells (experiments not shown), indicating that this protein band represents the ␣ 1 subunit of GABA A receptors. The weakly labeled lower molecular weight bands varied in labeling intensity in different experiments and could not be detected in untransfected HEK cells. They thus seemed to represent degradation products of the ␣ 1 subunit. The precipitation of ␣ 1 subunits by ␤ 3 (1-13) antibodies was not due to a cross-reactivity of these antibodies because it could not be observed in HEK cells transfected with ␣ 1 subunits only (experiment not shown).
When extracts from HEK cells co-transfected with ␣ 1 N and full-length ␤ 3 subunits or ␣ 1 N and ␤ 3 N constructs were precipitated with ␤ 3 (1-13) antibodies, three protein bands with apparent molecular masses 30, 33, and 36 kDa could be detected in Western blots using digoxygenized ␣ 1 (1-9) antibodies. The molecular mass of the smallest band (30 kDa) corresponds with that expected for the unglycosylated ␣ 1 N fragment. Because two glycosylation sites are present in the ␣ 1 subunit (29), the 33-and 36-kDa bands presumably represent partially and fully glycosylated ␣ 1 N fragments. The co-precipitation by ␤ 3 (1-13) antibodies of ␣ 1 N or ␣ 1 subunits indicates that not only fulllenth ␣ 1 and ␤ 3 subunits but also ␣ 1 N and ␤ 3 , ␣ 1 and ␤ 3 N, and even ␣ 1 N and ␤ 3 N constructs are able to form hetero-oligomers.
The subcellular distribution of these subunit combinations in HEK cells was investigated by immunofluorescence and confocal laser microscopy. Double staining with ␣ 1 (1-9) and the ␤ 2 /␤ 3 subunit-specific bd17 antibodies of intact HEK cells transfected with full-length ␣ 1 and ␤ 3 subunits (Fig. 5, A and B) FIG. 3. Sucrose density gradient centrifugation of GABA A receptors from the brain of young rats. Extracts from the brain of 8 -10-day-old rats were analyzed as described in Fig. 1. demonstrated that these subunits formed receptors expressed on the cell surface. Permeabilization of the cells indicated the additional presence of a large number of intracellular subunits with an identical subcellular distribution (Fig. 5, C and D). In HEK cells transfected with ␣ 1 N constructs and ␤ 3 subunits, only ␤ 3 subunits could be detected on the cell surface (Fig. 5, E  and F). These results are in agreement with previous reports demonstrating that ␤ 3 subunits are able to form homo-oligomeric receptors that are expressed on the cell surface (21,30). The observation that the truncated and the full-length subunit could be detected in the permeabilized cells in the same subcellular compartments (Fig. 5, G and H), indicates that ␤ 3 subunits that assembled with truncated ␣ 1 subunits were retained within the cell. No cell surface labeling was observed when HEK cells were co-transfected with ␣ 1 and ␤ 3 N or ␣ 1 N and ␤ 3 N constructs (Fig. 5, I, J, M, and N). However, ␣ 1 and ␤ 3 N (Fig. 5, K and L) or ␣ 1 N and ␤ 3 N subunits could be localized in the same subcellular compartments (Fig. 5, O and P).
To investigate whether assembly products from full-length and truncated subunits are able to form specific [ 3 H]muscimolbinding sites, membranes from nontransfected HEK cells or from cells transfected with ␣ 1 and ␤ 3 , ␣ 1 N and ␤ 3 , ␣ 1 and ␤ 3 N, or ␣ 1 N and ␤ 3 N were incubated with 5 nM of [ 3 H]muscimol in the absence or presence of 10 M GABA. For HEK cells transfected with ␣ 1 and ␤ 3 subunits, a specific [ 3 H]muscimol binding of 328 Ϯ 30 fmol/mg protein was found (Table I), whereas in cells co-transfected with ␣ 1 and ␤ 3 N constructs, a specific [ 3 H]muscimol binding of 21 Ϯ 4 fmol/mg protein was detected (Table I). In nontransfected HEK cells (not shown), however, and in cells co-transfected with ␣ 1 N and ␤ 3 or with ␣ 1 N and ␤ 3 N, no specific [ 3 H]muscimol binding could be identified. Scatchard analysis of equilibrium binding data indicated a high affinity [ 3 H]muscimol binding to HEK cells co-transfected with ␣ 1 and ␤ 3 N constructs (K D of 12.1 Ϯ 4.1 nM, B max of 78 Ϯ 29 fmol/mg protein, mean Ϯ S.E., n ϭ 4), and to cells transfected with ␣ 1 and ␤ 3 subunits (K D of 7.9 Ϯ 3.2 nM, B max of 805 Ϯ 53 fmol/mg protein, mean Ϯ S.E., n ϭ 4). Whereas the affinity for [ 3 H]muscimol of cells transfected with ␣ 1 and ␤ 3 N constructs or with ␣ 1 and ␤ 3 subunits was comparable (p ϭ 0.45, unpaired Student's t test), the B max values were significantly different (p Ͻ 0.0001, unpaired Student's t test). These results indicate that even intracellular and incomplete assembly intermediates can form specific high affinity [ 3 H]muscimol-binding sites. For the formation of this binding site, however, a full-length ␣ 1 subunit is necessary.
Density gradient centrifugation of constructs formed after transfection of HEK cells with ␣ 1 and ␤ 3 N combinations indicated broad peaks at 5.0 and 6.1 s. Because dimers composed of full-length subunits sediment at 5.5 s and trimers at 6.7 s, these data are compatible with the formation ␣ 1 ␤ 3 N dimers and trimers (Fig. 6). The lower sedimentation coefficients might have been due to the lower molecular mass of the truncated ␤ 3 N construct. The broad peak at 6.1 s might have been due to the formation of a mixture of intermediates composed of (␣ 1 ) 2 ␤ 3 N and ␣ 1 (␤ 3 N) 2 subunits.
Formation of the Benzodiazepine-binding Site-Because the benzodiazepine-binding site on GABA A receptors is located at the interface of ␣ 1 and ␥ 2 subunits (15), it was interesting to investigate whether this site could already be formed by ␣ 1 ␥ 2 dimers. Previous studies have indicated that HEK cells transfected with ␣ 1 and ␥ 2 subunits form high affinity [ 3 H]flunitrazepam-binding sites (23, 31) although predominantly forming subunit dimers (12). But the formation of minor amounts of higher oligomers and even completely assembled subunit pentamers could not be excluded by these studies.
To eliminate the possibility of formation of completely assembled subunit pentamers, in addition to the truncated ␣ 1 N construct a truncated ␥ 2 subunit (␥ 2 N) was cloned that again contained the complete extracellular N-terminal domain but no transmembrane domains. HEK cells were then co-transfected either with ␣ 1 and ␥ 2 subunits, ␣ 1 and ␥ 2 N, ␣ 1 N and ␥ 2 , or ␣ 1 N and ␥ 2 N subunits. Expressed subunits were extracted from these cells and were immunoprecipitated with ␥ 2 (1-33) antibodies. As shown in Fig. 4B, the full-length ␣ 1 or the truncated ␣ 1 N construct could be co-precipitated by ␥ 2 (1-33) antibodies from extracts of the appropriately co-transfected HEK cells. This was not due to a cross-reactivity of the ␥ 2 (1-33) antibody because this antibody (in contrast to ␣ 1 (1-9) antibodies) could not precipitate ␣ 1 subunits from HEK cells transfected with ␣ 1 subunits only (experiments not shown). These results therefore indicate that not only full-lenth ␣ 1 and ␥ 2 subunits but also ␣ 1 N and ␥ 2 , ␣ 1 and ␥ 2 N, and even ␣ 1 N and ␥ 2 N are able to form hetero-oligomers.
a For the combination ␣ 1 N␤ 3 only ␤ 3 subunits could be detected on the cell surface.
To investigate whether the structures formed from ␣ 1 and ␥ 2 , ␣ 1 N and ␥ 2 , ␣ 1 and ␥ 2 N, or ␣ 1 N and ␥ 2 N subunits were transported to the cell surface, appropriately transfected HEK cells were again investigated by immunofluorescence and confocal laser microscopy. As shown in Fig. 7 (A and B) for intact cells and in agreement with previous reports (4,21) no GABA A receptor subunits could be detected on the cell surface with ␣ 1 (1-9) or ␥ 2 (1-33) antibodies. After permeabilization of the cells, however, both subunits were detected in intracellular compartments (Fig. 7, C and D). For HEK cells co-transfected with ␣ 1 N and ␥ 2 , ␣ 1 and ␥ 2 N, or ␣ 1 N and ␥ 2 N, again no subunits could be detected on the cell surface. In permeabilized cells, however, a similar subcellular distribution of subunits was observed as in cells transfected with full-length ␣ 1 and ␥ 2 subunits (experiments not shown).
To investigate whether assembly products composed of fulllength and truncated or of two truncated subunits are able to form benzodiazepine-binding sites, membranes from nontransfected HEK cells or from cells co-transfected with ␣ 1 and ␥ 2 , ␣ 1 N and ␥ 2 , ␣ 1 and ␥ 2 N, or ␣ 1 N and ␥ 2 N were incubated with 5 nM  (Table II). These results indicate that not only full-length ␣ 1 and ␥ 2 subunits (31) but also truncated ␣ 1 and ␥ 2 subunits lacking transmembrane domains are capable of forming a benzodiazepine-binding site. Interestingly, however, the total number of binding sites observed in the four preparations was small compared with that observed in ␣ 1 ␤ 3 ␥ 2 transfected HEK cells in parallel experiments (874 Ϯ 19 fmol/mg protein). To investigate whether this was due to a low affinity or a low number of binding sites formed, Scatchard analysis was performed. Cells transfected with ␣ 1 and ␥ 2 subunits exhibited a K D of 135 Ϯ 38 nM and a B max of 301 Ϯ 23 fmol/mg protein (mean Ϯ S.E., n ϭ 4). Similar values were obtained when cells were transfected with ␣ 1 N and ␥ 2 (K D of 124 Ϯ 44 nM, B max of 322 Ϯ 49 fmol/mg protein, mean Ϯ S.E., n ϭ 4), ␣ 1 and ␥ 2 N, or ␣ 1 N and ␥ 2 N (data not shown). K D and B max values of cells transfected with ␣ 1 N and ␥ 2 subunits were comparable with those of cells transfected with ␣ 1 and ␥ 2 subunits (p ϭ 0.82 and p ϭ 0.91, respectively) but were significantly different (p ϭ 0.02 and p ϭ 0.0003, respectively) from cells transfected with ␣ 1 ␤ 3 ␥ 2 subunits (K D of 0.96 Ϯ 0.25 nM, B max of 1050 Ϯ 86 fmol/mg protein, mean Ϯ S. E., n ϭ 4).
Formation of the TPBS-binding Site-The TPBS-binding site of GABA A receptors can be identified in receptors composed of homo-oligomeric ␤ 3 subunits, ␣ 1 ␤ 3 and ␣ 1 ␤ 3 ␥ 2 subunits (23) and for the formation of this site the presence of the second transmembrane domain of the ␤ 3 subunit in a receptor is essential (18). It therefore was no surprise that only HEK cells transfected with ␣ 1 and ␤ 3 subunits but not those transfected with ␣ 1 and ␤ 3 N, or ␣ 1 N and ␤ 3 N subunits exhibited a specific [ 35 S]TBPS binding (experiments not shown). HEK cells transfected with ␣ 1 N and ␤ 3 subunits were not investigated for [ 35 S]TBPS binding because in these cells homo-oligomeric ␤ 3 receptors are formed (see above) that in any case exhibit high affinity [ 35 S]TBPS binding (23).
To investigate whether the TPBS-binding site can be formed by assembly intermediates containing ␣ 1 and ␤ 3 subunits, a ␤ 3 fragment was cloned (␤ 3 TM3) that not only contained the extracellular N-terminal domain but also the first three transmembrane domains of the ␤ 3 subunit. The ␤ 3 TM3 fragment could be co-precipitated with ␣ 1 subunits from HEK cells cotransfected with ␣ 1 and ␤ 3 TM3 (experiments not shown). Double staining of intact HEK cells transfected with full-length ␣ 1 and ␤ 3 TM3 constructs (Fig. 8, A and B) demonstrated that none of these subunits were expressed on the cell surface, but both subunits could be detected in the permeabilized cells in the same subcellular compartments (Fig. 8, C and D). However, no specific [ 35 S]TPBS binding could be identified in these cells.  7. Immunofluorescence of HEK cells co-transfected with full-length and truncated ␣ 1 and ␥ 2 subunits. HEK cells were transfected with ␣ 1 and ␥ 2 subunits. ␣ 1 subunits were labeled on the cell surface (A) or in permeabilized cells (C) using ␣ 1 (1-9) antibodies. ␥ 2 subunits were labeled on the cell surface (B) or in permeabilized cells (D) using ␥ 2 (1-33) antibodies. Rabbit antibodies were detected using anti-rabbit IgG Bodipy FL antibodies. Immunofluorescence was investigated by confocal laser microscopy (single sections). The experiment was performed five times with similar results.  ) was not significantly different from that of cells transfected with ␣ 1 and ␤ 3 subunits (7.9 Ϯ 3.2 nM), no change in the K D was to be expected in cells transfected with ␣ 1 ␤ 3 TM3. The increase in [ 3 H]muscimol binding, thus, presumably was due to an increase in the number of binding sites. This could have been caused by an increased stabilization of the [ 3 H]muscimol-binding site because of the presence of the three ␤ 3 transmembrane domains in the assembly product of ␣ 1 and ␤ 3 TM3 or by the formation of a second muscimol-binding site in a possible assembly product composed of two ␣ 1 and two ␤ 3 TM3 subunits.

Different Subunit Dimers Are Formed during GABA A Receptor
Assembly-The present study aimed to detect subunits or subunit combinations that could form the starting point of the GABA A receptor assembly process. However, neither in the adult rat brain nor in HEK cells transfected with ␣ 1 ␤ 3 ␥ 2 subunits and kept under standard tissue culture conditions could assembly intermediates be identified by density gradient centrifugation. This indicated that receptor synthesis in these tissues is either low and/or assembly of receptors is too fast to allow intermediates to be identified. When protein folding and subunit oligomerization of recombinant GABA A receptors was slowed down by reducing the culture temperature to 25°C, however, subunit monomers, dimers, trimers, tetramers, and pentamers could be detected by sucrose density gradient centrifugation. Interestingly, ␣ 1 , ␤ 3 , as well as ␥ 2 subunits could be identified in subunit dimers and all other oligomers. A similar result was obtained from brains of young rats, where a high expression of GABA A receptor subunits caused by ongoing development of the tissue leads to a constant high concentration of assembly intermediates. An identification of the subunit composition of the dimers was not possible, because the peaks for dimers and trimers were overlapping and could not be completely separated by density gradient centrifugation. A possible co-immunoprecipitation of two subunits in the dimer peak thus could have been caused by the respective subunit dimer or by a contamination with subunit trimers. In addition, the similarity of the apparent molecular masses of the ␣ 1 , ␤ 3 , and ␥ 2 subunits and the microheterogeneity of the labeled protein bands (12) prevented an identification of the exact dimers formed after radiolabeling of subunits by culturing with [ 35 S]methionin.
Although the formation of ␣ 1 ␤ 3 , ␣ 1 ␥ 2 , and ␤ 3 ␥ 2 heterodimers has been demonstrated previously in cells co-transfected with these subunit combinations (12), the presence of all three subunits in the dimer peak of brains from young rats or of cells transfected with ␣ 1 , ␤ 3 , and ␥ 2 subunits does not necessarily mean that all possible heterodimers are formed in these tissues. The data could also be explained by the formation of two different heterodimers or by the formation of heterodimers and/or homodimers. In addition, some of the dimers could be dead end products for the assembly of GABA A receptors and be subsequently degraded (5). The present data therefore cannot clarify the question of whether assembly of GABA A receptors can start from more than one possible dimer.
This question so far has also not been unequivocally answered for the nAChR. Thus, it has been reported that ␣ subunits of the nAChR first form heterodimers with ␥ and ␦, but not with ␤ subunits. The ␣␥ and ␣␦ heterodimers then were proposed to assemble with the ␤ subunit and with each other to form the complete ␣ 2 ␤␥␦ receptor (32). In another study, ␣␤␥ trimers were the first stable assembly intermediates identified (3), and it was proposed that the complete receptor is then formed by a stepwise addition of the ␦ and the second ␣ subunit. In any case, the complete assembly of the nAChR is a complex, slow, and inefficient process (33), and its mechanism is still not entirely clarified.

H]Muscimol but No [ 35 S]TBPS-binding Sites
Are Formed by ␣ 1 ␤ 3 Dimers and/or Trimers-Already during the early steps of assembly of the nAChR, subunit oligomerization and folding events lead to the formation of ligand-binding sites. Thus, a monomeric but properly folded ␣ subunit is sufficient for binding of the competitive antagonist ␣-bungarotoxin, whereas the formation of binding sites for agonists and low molecular weight antagonists occurs in ␣␥ and ␣␦ dimers (34). In the present study we therefore investigated the formation of ligand-binding sites on GABA A receptor intermediates.
In agreement with previous studies it was demonstrated that full-length ␣ 1 and ␤ 3 subunits on co-transfection into HEK cells form [ 3 H]muscimol as well as [ 35 S]TBPS-binding sites (23). These subunits are able to form pentameric receptors (12) and are expressed on the cell surface (4,21,30). C-terminally truncated ␣ 1 or ␤ 3 subunits, containing only the extracellular Nterminal domain also could assemble with each other or with full-length ␤ 3 or ␣ 1 subunits, respectively, but the assembly products remained in intracellular compartments and could not be detected on the cell surface. Specific [ 3 H]muscimol but no [ 35 S]TBPS-binding sites could be observed in HEK cells co-transfected with full-length ␣ 1 subunits and ␤ 3 N constructs. The absence of [ 35 S]TBPS-binding sites in these cells as well as in cells co-transfected with ␣ 1 N and ␤ 3 N subunits is not surprising, because recently it has been demonstrated that the presence of a TM2 region of the ␤ 3 subunit is essential for the formation of these sites (18).
Although similar amounts of subunits were expressed in ␣ 1 ␤ 3 N-or ␣ 1 ␤ 3 -transfected cells and although the affinity of [ 3 H]muscimol for the sites formed was comparable, the number of [ 3 H]muscimol-binding sites in ␣ 1 ␤ 3 N-transfected cells was small compared with that in ␣ 1 ␤ 3 -transfected cells. Immunoprecipitation studies and density gradient centrifugation indicated that most of the ␣ 1 subunits and ␤ 3 N fragments formed in the cells were assembled into heterodimers and heterotrimers, FIG. 8. Immunofluorescence of HEK cells co-transfected with full-length ␣ 1 and ␤ 3 TM3 constructs. Co-immunofluorescence was performed as described in the legend to Fig. 5. The experiments were performed three times with similar results. and only small amounts of these subunits remained unassembled. The comparatively small number of high affinity [ 3 H]muscimol-binding sites, thus, indicates that only a small part of the ␣ 1 ␤ 3 N heterodimers or heterotrimers formed contained these sites. This could have been due to a partially incorrect assembly of subunits or a low probability of formation of high affinity [ 3 H]muscimol-binding sites caused by the lacking transmembrane regions of the ␤ 3 N construct or the incomplete assembly of the receptor. The latter suggestion is supported by the finding that only a small number of unassembled nAChR ␣ subunits exhibited ␣-bungarotoxin binding but that the number of binding sites increased with additional assembly steps (3).
Because ␤ 3 subunits alone can form subunit pentamers exhibiting high affinity [ 35 S]TBPS-binding sites, a co-transfection with ␣ 1 N and ␤ 3 subunits could not be used to investigate whether [ 35 S]TBPS binding can be formed by assembly intermediates. Therefore, a ␤ 3 TM3 construct containing the N-terminal domain and three of the four transmembrane domains of the ␤ 3 subunit was transfected into HEK cells together with full-length ␣ 1 subunits. The observed absence of ␣ 1 subunits and ␤ 3 TM3 fragments on the cell surface suggests that neither ␣ 1 ␤ 3 TM3 hetero-nor ␤ 3 TM3 homo-pentamers were formed in these cells or that pentamers formed were retained within the cells. The finding that no [ 35 S]TBPS sites could be detected in ␣ 1 ␤ 3 TM3-transfected cells then either indicates that this site cannot be formed by an incompletely assembled receptor or that a complete ␤ 3 subunit in any case is essential for the correct formation of the [ 35 S]TBPS site.
[ 3 H]Ro 15-1788-binding Sites Are Formed on ␣ 1 ␥ 2 Dimers-In the present work we demonstrated that ␣ 1 and ␥ 2 , ␣ 1 N and ␥ 2 , ␣ 1 and ␥ 2 N, as well as ␣ 1 N and ␥ 2 N subunits are able to form hetero-oligomers that are not expressed on the cell surface but form specific [ 3 H]Ro 15-1788-binding sites. It has been reported previously that assembly of GABA A receptor ␣ 1 and ␥ 2 subunits on co-transfection into HEK cells predominantly stops at the stage of dimers (12). Because it is unprobable that higher oligomers are formed in the presence of truncated subunits, these results indicate that the formation of the benzodiazepine-binding site already occurs at the stage of heterodimers and that even intracellular and truncated ␣ 1 and ␥ 2 subunits lacking transmembrane domains are capable of binding benzodiazepines. The comparatively low affinity and number of the binding sites formed, however, indicates that [ 3 H]Ro 15-1788-binding sites formed by heterodimers do not significantly contribute to the total number of these binding sites formed in the brain.
Implications for the Function of GABA A Receptors-Although the [ 3 H]muscimol-binding site is formed by the N-terminal domain of the GABA A receptor ␣ and ␤ subunits (14), transmembrane domains seem also to support its formation. This is indicated by the observation that in HEK cells co-transfected with ␣ 1 N and ␤ 3 N constructs, in contrast to those transfected with ␣ 1 and ␤ 3 N constructs, no [ 3 H]muscimol-binding sites could be identified. Because binding of GABA to the [ 3 H]muscimol-binding site in intact GABA A receptors causes a conformational change in the transmembrane domains leading to the opening of the chloride ion channel (7), a close conformational interaction of the two subunit domains is to be expected. In addition, studies have indicated that point mutations within the second transmembrane domain (35) or the first extracellular loop between TM2 and TM3 (36) of subunits strongly influence gating of the channel.
Interestingly, however, ␣ 1 transmembrane domains seem to be more important than the corresponding ␤ 3 domains for the formation of a [ 3 H]muscimol-binding site. This is indicated by the observation that HEK cells transfected with full-length ␣ 1 subunits and ␤ 3 N constructs but not those transfected with full-length ␤ 3 subunits and ␣ 1 N constructs exhibit specific [ 3 H]muscimol-binding sites. Because ␣ subunits not only contribute to the formation of the [ 3 H]muscimol site (14) but also to the formation of the benzodiazepine-binding site (15), it is tempting to speculate that conformational changes in the chloride channel induced by GABA as well as the modulation of the GABA-induced current by benzodiazepines might predominantly be mediated by the ␣ subunit.
In contrast to the [ 3 H]muscimol-binding site that is expressed in cells transfected with ␣ 1 ␤ 3 or ␣ 1 ␤ 3 N combinations only, comparable amounts of [ 3 H]Ro 15-1788-binding sites are expressed in ␣ 1 ␥ 2 , ␣ 1 N␥ 2 , ␣ 1 ␥ 2 N, or ␣ 1 N␥ 2 N transfected cells. Binding affinity was similar for all these combinations but was more than 100-fold lower than that of ␣ 1 ␤ 3 ␥ 2 receptors, suggesting that the affinity of the [ 3 H]Ro 15-1788-binding site is influenced by the presence of additional subunits in a completely assembled receptor. Interestingly, however, the affinity of the benzodiazepine-binding site formed on subunit dimers is not influenced by the absence of ␣ 1 and ␥ 2 transmembrane domains, possibly reflecting an absence of a direct interaction between the benzodiazepine-binding site and the channel forming transmembrane domains. This conclusion is supported by the observation that binding of benzodiazepines does not cause direct opening of the chloride channel in the absence of GABA but enhances the frequency of GABA-induced channel opening (6). It thus can be speculated that binding of benzodiazepines to its site at the ␣␥ interface strongly influences the conformation of the GABA-binding site located at the other side (␣␤ interface) of the same ␣ subunit. This could either enhance the affinity of GABA for its binding site (6,7) or produce a conformational change similar to that produced by binding of GABA and thus reduce the number of GABA molecules necessary for opening the channel, as indicated by a previous study (37). Each of these mechanisms would enhance the frequency of channel opening by GABA. Further studies will have to decide between these possibilities.