Subunit arrangement of gamma-aminobutyric acid type A receptors.

The GABA(A) receptors are ligand-gated chloride channels. The subunit stoichiometry of the receptors is controversial; four, five, or six subunits per receptor molecule have been proposed for alphabeta receptors, whereas alphabetagamma receptors are assumed to be pentamers. In this study, alpha-beta and beta-alpha tandem cDNAs from the alpha1 and beta2 subunits of the GABA(A) receptor were constructed. We determined the minimal length of the linker that is required between the two subunits for functional channel expression for each of the tandem constructs. 10- and 23-amino acid residues are required for alpha-beta and beta-alpha, respectively. The tandem constructs either alone or in combination with each other failed to express functional channels in Xenopus oocytes. Therefore, we can exclude tetrameric or hexameric alphabeta GABA(A) receptors. We can also exclude proteolysis of the tandem constructs. In addition, the tandem constructs were combined with single alpha, beta, or gamma subunits to allow formation of pentameric arrangements. In contrast to the combination with alpha subunits, the combination with either beta or gamma subunits led to expression of functional channels. Therefore, a pentameric arrangement containing two alpha1 and three beta2 subunits is proposed for the receptor composed of alpha and beta subunits. Our findings also favor an arrangement betaalphagammabetaalpha for the receptor composed of alpha, beta, and gamma subunits.

to be pentameric (16 -18), the subunit stoichiometry of receptors composed of ␣ and ␤ is still controversial. Recombinantly expressed receptors have been reported as possibly tetrameric (19,20) as well as pentameric (18,21). Unitary dose-response curves for ␣␤ receptors, single IC 50 values for Zn 2ϩ inhibition, and unitary single channel properties (1) provide evidence against the formation of two populations of receptors, e.g. 2␣3␤ and 3␣2␤. A tetrameric rather than a pentameric structure has been proposed as one of several explanations for the lower average single channel conductance for the ␣␤ receptor as compared with the ␣␤␥ receptor (22,23).
A powerful way to gain insight into the arrangement of subunits in a multimeric channel is to predetermine the alignment of subunits by gene fusion and to analyze whether the linked subunits are able to form functional channels. This approach was first successfully applied to potassium channels (24 -26). Later it was also used to study subunit stoichiometry of other ion channels, e.g. a cyclic nucleotide-gated channel (27), the mechanosensitive channel MscL of Escherichia coli (28), and the cystic fibrosis conductance regulator channel (29). All these channels have their N and C termini on the cytoplasmic side so that the linkage occurs intracellularly. Up to now it has only been used once with limited success in the field of ligand-gated ion channels, which have both C and N termini on the extracellular side. Applying it to a GABA A receptor, Im et al. (30) prepared a tandem construct where the ␣6 subunit is linked to the ␤2 subunit via 10 glutamine residues and studied functional expression in HEK293 cells. The connection between the two subunits included the signal sequence of the ␤2 subunit of 24-amino acid residues in length. The consequences of such a signal sequence in the middle of a protein are difficult to predict.
We constructed here tandem constructs of ␣1 and ␤2 subunits for the first time in both arrangements ␣1-␤2 and ␤2-␣1. We determined the minimal length of the linkers necessary for the formation of functional channels. The constructs were expressed in Xenopus laevis oocytes either alone or in combination with single subunits to establish subunit stoichiometry and arrangement of GABA A receptors. We provide novel information on the architecture of GABA A receptors.

EXPERIMENTAL PROCEDURES
Construction of the Tandem cDNAs-Several ␣-␤ tandem cDNAs encoding a single polypeptide ␣␤ with linkers of differing length were established in the pCMV vector. The tandem constructs consisted of the modified rat ␣1 subunit (31) including its signal sequence at the N terminus and the mature rat ␤2 subunit at the C terminus. The modified ␣1 subunit differs from the original rat subunit by insertion of one amino acid residue. Insertion of this residue confers reactivity to the monoclonal antibody bd24 (32,31), which was essential for Western blot analysis (shown in the "Results" section). The ␣1 subunit was amplified by polymerase chain reaction using the pCHAI vector as template and the primers CATAGAAGACACCGGGACGA as a vector-specific primer and XTTGATGTGGTGTGGGGGCTTT as a gene-specific primer. The latter was complementary to the last codons before the stop codon and had the first part of the sequence coding for the respective linker attached (X). The ␤2 subunit was amplified using pCB2 as template and the primers ACTGACACACATTCCACAGCT as vector-specific primer and YCAGAGTGTCAATGACCCTAGT as a gene-specific primer. The latter was complementary to the first codons of the sequence of the mature protein and had the second part of the sequence coding for the respective linker attached (Y). The obtained fragments contained the open reading frame of the gene and some additional vector-derived sequence preceding or succeeding. The fragments were cut in the vector-derived sequence by EcoRI or XbaI, respectively, to be ligated in a three-fragment ligation into the pCMV vector cut with EcoRI and XbaI. The sequence of the resulting plasmids was verified. In the ␣-0-␤ tandem, the last amino acid residue of the ␣1 subunit is directly attached to the first amino acid residue of the mature ␤2 subunit. In the other tandems the following amino acid sequences are present between the N-terminal ␣1 subunit and the C-terminal ␤2 subunit: ␣-7-␤, Q 7 ; ␣-10-␤, Q 10 .
The ␤-␣ tandem cDNAs were prepared similarly. The ␤ subunit was amplified using CATAGAAGACACCGGGACGA as vector-specific and XGTTCACATAGTAAAGCCAATAGAC as gene-specific primer. The mature ␣ subunit was amplified using ACTGACACACATTCCACAGCT as vector-specific and YCAGCCGTCATTACAAGATGAA as gene-specific primer. The linkers introduced into the different ␤-␣ tandems are the following: ␤10␣, Q 10 ; ␤15␣, Q 5 A 3 PAQ 5 ; ␤20␣, Q 5 (A 3 P) 2 A 2 Q 5 ; ␤23␣, Q 3 (Q 2 A 3 PA) 2 AQ 5 . A long sequence of consecutive glutamine residues might exhaust the respective tRNA pool during protein synthesis and therefore lead to an early termination of the synthesized protein. Therefore, other amino acid residues were introduced. Alanine and proline residues were chosen for their properties to form no distinct secondary structure elements.
Two-electrode Voltage Clamp-All measurements were done in medium containing 5 mM HEPES, pH 7.4, 90 mM NaCl, 1 mM MgCl 2 , 1 mM KCl, and 1 mM CaCl 2 at a holding potential of Ϫ80 mV. For the determination of maximal current amplitudes 1 mM GABA (Fluka, Switzerland) was applied for 20 s. Sodium currents were determined by a potential jump from a holding potential of Ϫ100 to Ϫ15 mV (Fig. 1). The GABA-evoked peak current amplitude was standardized to the co-expressed sodium current amplitude of the same oocyte. The mean standardized current amplitude of at least three oocytes per subunit combination was then compared with the mean standardized wild type current amplitude. Current stimulation by diazepam was determined at a GABA concentration evoking 5% of the maximal current amplitude in combination with 1 M diazepam (Roche Molecular Biochemicals).
Western Blotting-Oocytes were homogenized in lysis buffer (10 mM HEPES, pH 8.0, 100 mM NaCl, 10 mM EDTA, 1% Triton X-100, pepstatin, leustatin, antipain, and phenylmethylsulfonyl fluoride, each at 5 g/ml) using a Teflon glass homogenizer. The homogenate was incubated on ice for 15 min and centrifuged at 15,000 ϫ g for 15 min at 4°C. The supernatant was subjected to SDS-polyacrylamide gel electrophoresis (35). Proteins were transferred to nitrocellulose membranes (HybondECL, Amersham Pharmacia Biotech) according to Towbin et al. (36) and decorated with the monoclonal antibody bd24 (31, 32), which recognizes the N terminus of the ␣1 subunit of the GABA A receptor.
Bands were detected using the ECL system (Amersham Pharmacia Biotech).

Preparation and Analysis of Tandem Constructs-Tandem
cDNAs were constructed that consisted of the ␣1 and the ␤2 subunit of the GABA A receptor in both arrangements, ␣1-␤2 and ␤2-␣1 ( Fig. 2A). The N-terminal ␣1 (␤2) subunit was taken in its precursor form to ensure insertion into the membrane mediated by the signal sequence. The C-terminal ␤2 (␣1) subunit was depleted of its signal sequence because it is difficult to predict what consequences this stretch of 24 (27) mostly hydrophobic amino acid residues in the middle of the new fusion protein will have. To bridge the distance between the C terminus of the ␣1 (␤2) subunit and the N terminus of the ␤2 (␣1) subunit, we introduced synthetic linkers of different length to determine the shortest possible linker resulting in a functional fusion protein after expression in the Xenopus oocyte. Although the subunits of the GABA A receptor share the same topology and have a high sequence homology, they differ slightly in the number of amino acid residues after the fourth predicted transmembrane region at the C terminus as well as at the beginning of the N-terminal portion of the subunit. Therefore, linkers of different length were tested for the ␣1-␤2 and ␤2-␣1 construct separately.
The function of the different tandem constructs was assessed after expression in Xenopus oocytes, either alone (Fig. 2B, I and II), in combination with each other (Fig. 2B, III), or in combination with single ␣1 (Fig. 2B, IV and V) or ␤2 (Fig. 1B, VI and VII) subunits. Expression of tandem constructs alone is predicted to yield receptors composed of an even number of subunits, whereas combination of tandem constructs with single subunits can additionally result in receptors with an uneven number of subunits.
The first criterion for normal channel function was a GABAevoked maximal current amplitude comparable with that of receptors made from single ␣1 and ␤2 subunits, the wild type receptor. These current amplitudes amounted to 1.5-8 A. Expression of either ␣1 or ␤2 subunits alone failed to produce detectable currents. To compensate for differences in the expression level between the individual oocytes, GABA-induced current amplitudes were standardized to the current amplitude of the co-expressed voltage-gated sodium channel in the same oocyte. Constructs were examined for standardized maximal current amplitudes (I max ) and the apparent affinity for GABA (K a ). Only receptors performing very similarly to the wild type receptor regarding I max and K a were considered fully functional. Oocytes were either expressing ␣ and ␤ subunits of GABA A receptors and voltage-gated sodium channels (␣/␤/Na) or single ␣ subunits (␣/Na) or ␤ subunits (␤/Na) together with voltage-gated sodium channels. The duration of the application of GABA or of the potential jump from Ϫ100 to Ϫ15 mV is shown above the respective traces.

The Tandem Constructs Are Not Proteolyzed in the Linker
Sequence-To evaluate whether the expressed tandem constructs were intact or subjected to proteolysis we analyzed the newly formed GABA A receptors by Western blotting. The monoclonal antibody bd24 against the N-terminal of the ␣1 subunit (31, 32) was used. Fig. 3 shows that single ␣1 subunits of wild type receptors migrate at 50 kDa (lane 1). This specific band is missing in the ␣-10-␤/␤ combination (lane 2), thus indicating the absence of monomeric ␣1 subunit and, therefore, of significant proteolysis of the linker. A very faint unspecific signal at the 50-kDa position is also seen for non-injected oocytes (lane 3). As indicated by a strong signal, the ␣-10-␤ tandem construct migrates at 120 -140 kDa. The absence of any additional band with bd24 reactivity also excludes proteolysis elsewhere in the construct. A peptide containing bd24 reactivity that is larger than 29 kDa would have been seen. The ␤-23-␣ tandem construct could not be detected because the epitope for the antibody seems to include the free N terminus of the ␣1 subunit, which is blocked by the linker in this construct. The small quantity of channel expressed prevented detection with another antibody due to insufficient sensitivity. As described below, we also have functional evidence for the fact that proteolysis of both tandem constructs can be excluded.
The Length of the Linker Is Critical for Functional Expression of Linked Subunits-If the ␣1-␤2 tandem construct was co-expressed with single ␤2 subunits, functional channels were formed provided the linker was long enough (Fig. 4). With no additional linker but only the 13 amino acid residues after the fourth transmembrane region of the ␣1 subunit (␣-0-␤) connected to the N terminus of the ␤2 subunit, no current was detectable in injected oocytes. With a linker of 7 residues in length (␣-7-␤), we found standardized maximal current amplitudes that remained below those expressed from wild type receptors, whereas the tandem construct with a linker of 10 residues (␣-10-␤) resulted in similar standardized maximal current amplitudes. The dose-response curves of the ␣-7-␤ and the ␣-10-␤ tandem constructs were close to that of the wild type receptors (Fig. 5A). The two constructs resulted in channels with similar K a values of 9 Ϯ 3 and 11 Ϯ 2 M, respectively, comparable with the combination of single ␣ and ␤ subunits with a K a of 9 Ϯ 2 M, pointing to an unchanged apparent affinity for GABA despite the covalent linkage.
On the right panel of Fig. 4 the results of the analogue examination for the ␤2-␣1 constructs are shown. There was almost no detectable current when we combined the constructs with linkers of 10-and 15-amino acid residues with single ␤2 FIG. 2. A, schematic drawing of the ␣-␤ tandem construct. The C terminus of the ␣1 subunit is linked to the N terminus of the ␤2 subunit by linkers of different length. B, theoretically possible subunit arrangements of ␣-␤ and ␤-␣ tandem constructs in a tetrameric (I-III) or a pentameric (IV-VII) receptor. Arrangements I and II are identical and can be formed by both tandem constructs ␣-␤ (I) or ␤-␣ (II). Arrangement III can be formed by one ␣-␤ and one ␤-␣ tandem construct. We assume the presence of at least two ␣ and two ␤ subunits in a pentameric receptor. Both tandem constructs ␣-␤ and ␤-␣ yield in this case the same arrangement when combined with a single ␣ subunit (IV and V) or a single ␤ subunit (VI and VII), respectively. subunits. A tandem construct with a linker of 20 residues produced receptors with standardized maximal current amplitudes similar to those of wild type receptors. However, the dose-response curve (Fig. 5B) was shifted to the right, i.e. the apparent affinity for GABA was reduced. With 64 Ϯ 33 M, the K a was about 7-fold higher than that of the wild type receptors. A tandem construct containing a linker of 23 residues also reached standardized maximal current amplitudes similar to wild type receptors. The GABA dose-response curve for these channels (Fig. 5B) is characterized by a K a of 20 Ϯ 2 M, which is close to the wild type receptor, with a K a of 9 Ϯ 2 M.
GABA A Receptors Made from ␣1 and ␤2 Subunits Are Pentamers Containing 2 ␣ and 3 ␤ Subunits-The two functional tandem constructs ␣-10-␤ and ␤-23-␣ were analyzed further. When either the ␣-10-␤ or the ␤-23-␣ constructs were expressed alone, we hardly detected GABA evoked currents (Fig. 6A). The co-expressed voltage-gated sodium channel showed the same expression levels in oocytes expressing tandem constructs or wild type receptors. Thus, the absence of RNase activity and the capability of protein expression in the individual oocyte was confirmed. Moreover we exclude proteolysis for either construct because proteolysis of the linker would in each case liberate ␣1 and ␤2 subunits, which in turn should result in functional channels. When ␣-10-␤ and ␤-23-␣ constructs were expressed in the same oocyte, the standardized maximal current amplitudes remained below 10% of the wild type current (Fig. 6C). These results led to the conclusion that tetrameric receptors of the arrangement ␣␤␣␤, which is equal to the arrangement ␤␣␤␣ (see Fig. 2B, I and II) or of the arrangement ␣␤␤␣ (see Fig. 2B, III) do not correspond to a functional receptor made from single ␣ and ␤ subunits.
The tandem constructs ␣-10-␤ and ␤-23-␣ were also coex-pressed with single ␣1 subunits and analyzed for maximal current amplitudes. Almost no current was detected upon application of GABA (Fig. 6B), whereas sodium currents were expressed in the same oocytes. The addition of a single ␣1 subunit to the combination of both tandem constructs ␣-10-␤ and ␤-23-␣ resulted in slightly elevated maximal current amplitudes as compared with the combination of ␣-10-␤ with ␤-23-␣ (Fig. 6C), but they were still far below those of the wild type receptors. This indicates an inefficient formation of functional channels in this case.
FIG. 5. Dose-response curves of tandem constructs co-expressed with single ␤2 subunits. A, the ␣-␤ tandem constructs with linkers of 7-or 10-amino acid residues in length resulted in channels with an unchanged apparent affinity for GABA. B, channels from the ␤-␣ tandem construct with a linker of 20-amino acid residues show a slightly reduced apparent affinity for GABA, whereas for channels from the construct with a linker of 23 amino acid residues, the apparent affinity is close to the one of the combination of single ␣ and ␤ subunits.  Fig. 4 and 6A. C, proposed rearrangement of subunits in a tandem construct. The marked areas in the schematic subunits (stripes in ␥, points in ␣) represent amino acid residues that can contribute to the benzodiazepine binding site. Note that a proper benzodiazepine binding site at a ␣␥ subunit interface can only be formed in one of the two different subunit arrangements (II) and is lost in the other (I). Fig. 6D shows that both the ␣-10-␤ and the ␤-23-␣ tandem constructs could be complemented with single ␤2 subunits to form functional channels. This result matches the theoretical consideration that both tandem constructs yield the same arrangement when complemented with a single ␤2 subunit (compare Fig. 2B, VI and VII).
Coexpression of the Tandem Constructs with a Single ␥2 Subunit-When the ␣-10-␤ tandem construct is complemented with a single ␥2 subunit, the standardized maximal current amplitude amounts to about 26% compared with the wild type receptor (Fig. 6E). Submaximal current amplitudes can be stimulated by diazepam by 134 Ϯ 8% (mean Ϯ S.D., n ϭ 3) (not shown). The ␤-23-␣ tandem construct complemented with a single ␥ subunit results in functional channels with standardized maximal current amplitudes similar to wild type receptors (Fig. 6E). Submaximal current amplitudes of these receptors are also stimulated by diazepam by 360 Ϯ 10% (mean Ϯ S.D., n ϭ 3) (not shown).

DISCUSSION
In this study we have demonstrated the feasibility of covalent subunit linkage ␣1-␤2 and ␤2-␣1 for the GABA A receptor channel. We have also established the minimal linker lengths required for functional expression. Our results strongly suggest a pentameric structure of the GABA A receptor composed of ␣1 and ␤2 subunits and exclude a tetramer. The technique described here may also be applied to the study of other ligandgated ion channels.
Tandem linkage of subunits is a powerful strategy to extract information about stoichiometry and arrangement of multimeric proteins. This approach has first been applied to the study of potassium channels (24). Later, Im et al. (30) made a tandem construct consisting of the GABA A receptor subunit precursors ␣6 and ␤2. They found that their ␣6-␤2 tandem construct alone failed to produce functional GABA channels, but combination with either single ␣6 or ␥2 subunits, but not ␤2 subunits, restored receptor function after expression in HEK293 cells. Functional expression was, however, very low in all these cases and did not exceed 0.2 nA even for the wild type subunit combination ␣6 and ␤2 (30).
In the present tandem constructs we omitted the signal sequence stretch of the second subunit, which might have unpredictable effects on e.g. protein folding, insertion of the protein into the membrane, subunit assembly, or proteolysis of the connection between the subunits. We linked the ␣1 and the ␤2 subunits of the GABA A receptor in both arrangements and expressed the resulting tandem constructs ␣-␤ and ␤-␣ in Xenopus oocytes. They were both shown to result in functional channels when complemented with ␤2 subunits. When the tandem constructs were expressed either alone or in combination with each other, no functional receptors were formed. Therefore, our most important conclusion here is that the GABA A receptor made from ␣1 and ␤2 subunits is not composed of an even number of subunits. We can exclude tetrameric receptors of the subunit arrangements ␣␤␣␤ from the expression of each of the tandem constructs alone and the arrangement ␣␤␤␣ from their co-expression. Only arrange-ments of a 1:1 stoichiometry of ␣ and ␤ subunits have been tested here because stoichiometries for ␣␤ receptors of 3:1 or 1:3 have been shown to be unlikely (19,20). These findings confirm the conclusion drawn from Western blot analysis that proteolytic cleavage in the sequence of the linker (Fig. 7A) does not occur to a significant extent. The participation of only one subunit of the tandem construct in the functional receptor (Fig.  7B) can also be excluded. If either one or both of these events had occurred, the formation of functional pentameric receptors from the tandem constructs alone would have been observed.
The observation that both tandem constructs form functional channels in combination with single ␤2 subunits but fail to do so in combination with single ␣1 subunits supports the view that a receptor made from ␣ and ␤ subunits is a pentamer composed of two ␣ and three ␤ subunits. This had also been proposed based on immunoprecipitation experiments in HEK293 cells expressing ␣1␤3 receptors (18). A receptor stoichiometry of three ␣6 and two ␤2 subunits has also been suggested (30). This might indicate that the subunit stoichiometry of an ␣␤ receptor depends on the specific subunit isoforms expressed together and/or on differences in the expression systems used.
A further aim of this study was the design of optimal linkers between the subunits. The linkage of two subunits should position both next to each other in the receptor. When no functional channels can be detected, the forced neighborhood of the two subunits either prohibits proper channel formation, or the linker is too short. When, in contrast, functional channels can be expressed from linked subunits, their neighborhood may be assumed unless the linker is very long. In this case the two linked subunits do not necessarily locate next to each other in the receptor multimer. It is then possible for another subunit to position itself between the two linked subunits. We therefore determined the minimal linker length for both, the ␣-␤ and the ␤-␣ tandem constructs, necessary for the formation of functional channels. We found this length to be 10 and 23 amino acid residues, respectively. Shorter linkers altered the apparent affinity for GABA or the maximal current amplitude of the channel, probably by distorting the conformation of the resulting receptor. It should be noted that the ␣-7-␤ and ␤-20-␣ tandem constructs, which have linkers that are 3 amino acid residues or about 11 Å shorter, performed nearly as well as wild type receptors. Therefore, the optimal linker length may be somewhat shorter than 10 or 23 amino acid residues, respectively. In our calculation of the actual linker length we included the synthetic linker as well as the C-and N-terminal elongations of the respective subunits (Table I). We assumed an extended conformation of both with 3.6-Å per amino acid residue. In this case the total length of the subunit connection may be estimated to be maximally 83 and 97 Å in the ␣-␤ and the ␤-␣ tandem construct, respectively, which might be diminished by the existence of secondary structure elements. For the reasons mentioned above, we assume that the actual linker length is substantially shorter. It is of interest to estimate whether these respective linker lengths allow interspersing of an additional subunit. We can consider the nicotinic acetylcholine receptor an appropriate model for the structure of the GABA A Total number of residues 23 27 receptor, as they both belong to the same superfamily of ligandgated ion channels. The three-dimensional structure of the nicotinic acetylcholine receptor has been resolved to 4.6 Å (37). All the members of the superfamily share a high sequence homology and the same topology, and it is assumed that they also have a very similar overall shape. From the dimensions of the receptor we can estimate the minimal length of a peptide passing along the perimeter of one subunit to be about at least 54 Å if the N terminus is located at the membrane surface. This minimal length of 54 Å is unrealistic for the following reasons. First, the receptor surface is certainly not smooth, but irregular. Second, the N terminus of the second subunit of the tandem construct is not necessarily located at the membrane surface as the beginning of the connection is predicted to be. Most importantly, location of either the N terminus or the C terminus away from the opposed edges of the linked subunits would both result in a corresponding increase of the required minimal length. Comparing the maximal length of the subunit connections and the minimal length such a connection must have to surround an additional subunit and the restrictions made to these values, we consider it unlikely that another subunit is interspersing, but we cannot entirely exclude this possibility.
In initial experiments we combined the two tandem constructs ␣-10-␤ and ␤-23-␣ each with single ␥ subunits. In the case of the ␤-23-␣ tandem construct, the resulting channel exhibited the same maximal current amplitude as wild type receptors, whereas in the case of the ␣-10-␤ tandem construct, maximal current amplitudes remained below that of wild type receptors. The fact that both tandem constructs ␣-␤ and ␤-␣ resulted in channels sensitive to diazepam was very surprising. The binding site for benzodiazepines is thought to be located at the ␣␥ subunit interface (38). This defined interface is lost in one of the two arrangements I and II shown in Fig. 7C. It is possible that the ␤␥ subunit interface can take over benzodiazepine binding properties, as it has been observed that receptors expressed from only ␤ and ␥ subunits are sensitive to benzodiazepines (39). An alternative and more likely interpretation is based on a rearrangement of one of the tandem constructs. We suggest this rearrangement for the following reason. In the presence of ␥ subunits, receptors containing ␣ and ␤ subunits alone are no more formed (23), but the ␥ subunit seems to induce a subunit assembly leading to ␣␤␥ receptors (40,18). The assembly of the tandem constructs with the ␥ subunits might, thus, start with the formation of proper ␣␥ or ␥␤ subunit interfaces. Then the second subunit of the tandem would be integrated. In the case of the ␤-23-␣ tandem construct this happens very efficiently, resulting in channels with current amplitudes similar to wild type receptors. In contrast, the ␣-10-␤ tandem constructs have to reorient to adopt a ␤-␣ arrangement (Fig. 7C, II). The linker might now be too short and disturb the proper conformation of the subunits. It is also conceivable that the rearrangement proceeds inefficiently. Therefore, the maximal current amplitude is lower; nevertheless the proper binding sites for GABA and benzodiazepines, both, seem to be present. Thus, we propose the subunit arrangement ␤␣␥␤␣ for the ␣1␤2␥2 receptor.
A rearrangement as proposed for the tandem construct ␣-␤ in the presence of a ␥2 subunit would not result in additional subunit arrangements in the case of a tetrameric receptor but would add another possible subunit sequence in pentameric receptors composed of only ␣ and ␤ subunits. If one of the tandem constructs in Fig. 2B, VI, reorients, an arrangement, ␣␣␤␤␤, which is not shown, will be formed. This additional arrangement can not be excluded from our data.
The preparation of triple constructs containing the ␥ subunit and its co-expression with the ␤-␣ tandem construct will allow the study of the effect of single point mutations exclusively in one defined ␣ or ␤ subunit. For topological reasons it can be safely predicted that subunits linked in a triple construct are not able to rearrange. It will also be possible to study the positional effect of different subunit isoforms in the same receptor pentamer.
In summary, we have demonstrated the feasibility of covalent subunit linkage for a ligand-gated ion channel. For the first time we have established the minimal linker lengths required for functional expression. Our results strongly suggest a pentameric structure of the ␣1␤2 GABA A receptor and exclude a tetramer. This work provides a new perspective for the study of subunit arrangement also of other ligand-gated ion channels.