Unanticipated Structural and Functional Properties of δ-Subunit-containing GABAA Receptors*

GABAA receptors mediate inhibitory neurotransmission in the mammalian brain via synaptic and extrasynaptic receptors. The delta (δ)-subunit-containing receptors are expressed exclusively extra-synaptically and mediate tonic inhibition. In the present study, we were interested in determining the architecture of receptors containing the δ-subunit. To investigate this, we predefined the subunit arrangement by concatenation. We prepared five dual and three triple concatenated subunit constructs. These concatenated dual and triple constructs were used to predefine nine different GABAA receptor pentamers. These pentamers composed of α1-, β3-, and δ-subunits were expressed in Xenopus oocytes and maximal currents elicited in response to 1 mm GABA were determined in the presence and absence of THDOC (3α, 21-dihydroxy-5α-pregnane-20-one). β3-α1-δ/α1-β3 and β3-α1-δ/β3-α1 resulted in the expression of large currents in response to GABA. Interestingly, the presence of the neurosteroid THDOC uncovered α1-β3-α1/β3-δ receptors, additionally. The functional receptors were characterized in detail using the agonist GABA, THDOC, Zn2+, and ethanol and their properties were compared with those of non-concatenated α1β3 and α1β3δ receptors. Each concatenated receptor isoform displayed a specific set of properties, but none of them responded to 30 mm ethanol. We conclude from the investigated receptors that δ can assume multiple positions in the receptor pentamer. The GABA dose-response properties of α1-β3-α1/β3-δ and β3-α1-δ/α1-β3 match most closely the properties of non-concatenated α1β3δ receptors. Furthermore, we show that the δ-subunit can contribute to the formation of an agonist site in α1-β3-α1/β3-δ receptors.

In the absence of any method able to determine membrane protein architecture in situ in the nervous system, model systems have to be used. Recently, ␣␤␣␦␤ has been proposed to be the predominant subunit arrangement around the pore when viewed from the extracellular space for ␣ 4 ␤ 3 ␦ GABA A receptors expressed in tsA 201 cells (40). This work was performed at the structural level. In the present study, we have focused on the architecture of ␣ 1 ␤ 3 ␦ GABA A receptors expressed in Xenopus oocytes at the functional level. To investigate active channels, we used covalently linked ␣ 1 , ␤ 3 , and ␦ subunits to have a defined arrangement of different subunits in a pentamer (41)(42)(43)(44)(45)(46). The concatenated receptors were characterized in detail using the agonist GABA, the neurosteroid THDOC, Zn 2ϩ , and ethanol, and their properties were compared with those of non-of the RNA with known concentrations of RNA ladder (GIBCO Invitrogen) as standard on the same gel. The cRNAs were dissolved in water and stored at Ϫ80°C. Isolation of oocytes from the frogs, culturing of the oocytes, injection of cRNA, and defolliculation were done as described earlier (47). cRNA coding for each dual and triple subunit concatemer was injected either alone or in different combinations in oocytes resulting in a total of seven different concatenated receptors. Oocytes were injected with 50 nl of RNA solution containing each dual or triple subunit construct at 50 nM and pentameric constructs at 100 nM, unless indicated otherwise in Fig. 1. Combinations of ␣ 1 -, ␤ 3 -, and ␦-subunits were expressed at a ratio of 10:10:10 nM or 10:10:50 nM. If the ␥ 2 -subunit is used in place of ␦, the latter ratio is required (48). Potentiation of currents by THDOC was similar, but current amplitudes were rather small in the former case (data not shown). Therefore, we used the second condition for detailed characterization. The injected oocytes were incubated in modified Barth's solution (47) at 18°C for about 72 h for the determination of I max and for at least 24 h before the measurements for detailed characterization of the functional receptors.
Two-electrode Voltage Clamp Measurements-All measurements were done in medium containing 90 mM NaCl, 1 mM MgCl 2 , 1 mM KCl, 1 mM CaCl 2 , and 5 mM HEPES pH 7.4 at a holding potential of Ϫ80 mV. For the determination of maximal current amplitudes 1 mM GABA (Fluka) was applied in the absence and presence of 1 M THDOC (Sigma) for 20 s. THDOC was prepared as a 10 mM stock solution in dimethyl sulfoxide (DMSO) and was dissolved in external solution resulting in a maximal final DMSO concentration of 0.5%. The perfusion solution (6 ml/min) was applied through a glass capillary with an inner diameter of 1.35 mm, the mouth of which was placed about 0.4 mm from the surface of the oocyte (5). Non-concatenated and concatenated receptors containing the ␦-subunit showed a pronounced decrease in response to GABA with time. This decrease amounted to about 30 -70% and did not recover. The experiments were performed after the measured currents became constant. Concentration response curves for GABA were fitted with the equation I(c) ϭ I max /(1 ϩ (EC 50 /c) n ), where c is the concentration of GABA, EC 50 the concentration of GABA eliciting half maximal current amplitude, I max is the maximal current amplitude, I the current amplitude, and n the Hill coefficient.
Relative current potentiation by THDOC was determined as (I 1 M THDOC ϩ 1 mM GABA /I 1 mM GABA Ϫ 1) ϫ 100%. Inhibition curves for Zn 2ϩ were fitted with the equation I(c) ϭ I(0)/(1 ϩ (IC 50 /c) n ), where I(0) is the control current in the absence of Zn 2ϩ standardized to 100%, I is the relative current amplitude, c is the concentration of Zn 2ϩ , IC 50 the concentration of Zn 2ϩ causing 50% inhibition of the current, and n the Hill coefficient. Zn 2ϩ was pre-applied for a minimum of 1 min prior to co-application of GABA with Zn 2ϩ . Potentiation by ethanol was determined at EC 20 for GABA, using 30 mM ethanol. Relative current potentiation by ethanol was determined as (I 30 M ethanol ϩ GABA EC20 /I GABA EC20 Ϫ 1) ϫ 100%.
Data are given as mean Ϯ S.E. for the I max values for GABA with and without THDOC and as mean Ϯ S.D. for analysis of properties of receptors using GABA, Zn 2ϩ , and ethanol. The perfusion system was cleaned between two experiments by ␦-Subunit-containing GABA A Receptors washing with 100% dimethyl sulfoxide (DMSO) after application of THDOC and with 10 mM HCl for Zn 2ϩ experiments to avoid contamination.

RESULTS
Preparation of Concatenated ␦-Subunit-containing GABA A Receptors-We used the subunit concatenation approach to determine the architecture of ␦-subunit-containing GABA A receptors. We assumed that the ␦-subunit would either occupy the position of the ␥ 2 -subunit, one of the two ␣-subunits, or one of the two ␤-subunits in the major isoform of GABA A receptors that is arranged ␥ 2 ␤ 2 ␣ 1 ␤ 2 ␣ 1 counter-clockwise when viewed from the synaptic cleft (41,42). We also analyzed receptors containing two ␦-subunits in the same receptor with ␦ at ␥ position and one of the ␤ positions. Five dual and three triple concatenated constructs were prepared to force the ␦-subunit into defined positions to form nine different GABA A receptor pentamers (Fig. 1). For the design of the linkers, we applied the rule that the sum of the predicted C-terminal protrusion of a preceding subunit and the artificial linker has to be minimally 23 residues in length. Shorter linkers do not result in receptor expression (41,46).
Functional Expression of ␦-Subunit-containing GABA A Receptors-Concatenated receptors R1-R9 ( Fig. 1) were expressed in Xenopus oocytes. The non-concatenated subunit combinations ␣ 1 ␦, ␤ 3 ␦, ␣ 1 ␤ 3 , ␣ 1 ␤ 3 ␦ and concatenated dual and triple subunit constructs were used as a control. GABA has been shown to be a partial agonist for ␦-subunit-containing receptors (31,32), and the maximal current evoked by GABA could be enhanced by a neurosteroid. Here we estimated receptor expression in the presence of the neurosteroid THDOC. Currents were determined at saturating concentration of GABA (1 mM) in the absence and presence of 1 M THDOC (Fig. 1). The non-concatenated ␣ 1 ␦ and ␤ 3 ␦ receptors resulted in currents Ͻ 10 nA in either case. Both, ␣ 1 -and ␤ 3 -subunits were required to obtain robust expression of ␦-subunit-containing receptors (Fig. 1). Further, non-concatenated ␣ 1 ␤ 3 receptors were expressed to compare their properties with those of ␣ 1 ␤ 3 ␦ receptors to ensure that ␦-subunit was being expressed in the latter receptors. ␣ 1 ␤ 3 and ␣ 1 ␤ 3 ␦ displayed a different sensitivity toward THDOC. Current potentiation was 6-and 17-fold, respectively. This difference, together with the differential sensitivity to Zn 2ϩ (see below), confirms that ␦-subunit was indeed being incorporated into ␣ 1 ␤ 3 ␦ receptors, although we cannot completely rule out that a subpopulation of ␣ 1 ␤ 3 receptors is expressed along with ␣ 1 ␤ 3 ␦. To our surprise and for reasons we can only speculate (see "Discussion") the concatenated ␤ 3 -␣ 1 construct in the absence and presence of THDOC and the ␤ 3 -␣ 1 -␦ construct in the presence of THDOC, themselves, resulted in substantial current expression, unlike all the other concatenated subunits (Fig. 1). For the ␤ 3 -␣ 1 construct this functional expression was analyzed further. When 50 nM construct were injected, the current was about 500 nA. Expression of 25 nM of ␤ 3 -␣ 1 resulted in about 30% of this current, and 10 nM gave currents less than 7% of the currents observed using 50 nM of cRNA ( Fig. 1). So clearly this artifactual signal was only prominent at higher cRNA concentrations used. Unfortunately, high concentrations of cRNA were required for the ␦-subunit-containing receptors to achieve significant expression. To exclude ambiguities in the interpretation of the properties observed for the receptors ␤ 3 -␣ 1 -␦/␣ 1 -␤ 3 (R1) and ␤ 3 -␣ 1 -␦/␤ 3 -␣ 1 (R5), we prepared the ␤ 3 -␣ 1 -␦-␣ 1 -␤ 3 (P1) and ␤ 3 -␣ 1 -␦-␤ 3 -␣ 1 (P5) pentamers. Notably, receptors containing the ␦-subunit in different positions resulted in the current expression. The concatenated receptors with the subunit arrangement , and ␤ 3 -␣ 1 -␦/␦-␣ 1 (R9) resulted in currents Ͻ45 nA on co-application of GABA and THDOC. None of the functional receptors was directly activated by 1 M THDOC alone (data not shown). With the exception of ␤ 3 -␣ 1 -

␦-Subunit-containing GABA A Receptors
␦/␣ 1 -␤ 3 (R1), THDOC significantly potentiated the maximal current amplitudes elicited by GABA. ␣ 1 -␤ 3 -␣ 1 /␤ 3 -␦ (R2) is especially remarkable in this respect as its expression was not strongly evident with GABA alone and was only uncovered in the presence of THDOC. Potentiation of maximal currents by THDOC amounted to about 22-fold in this case.
It was interesting to see if it was possible to place ␦ in one of the ␣ positions. While ␤ 3 -␦-␤ 3 /␣ 1 -␤ 3 (R6) did not result in current expression, ␤ 3 -␦-␤ 3 /␤ 3 -␣ 1 (R7) resulted in currents of slightly larger amplitude as ␤ 3 -␣ 1 . The current showed likewise relatively little stimulation by THDOC. The EC 50 was determined as 40 Ϯ 18 M and the Hill coefficient as 0.8 Ϯ 0.1 (n ϭ 5). Again these parameters are reminiscent of the current mediated by ␤ 3 -␣ 1 . Therefore, we assume that R7 is probably not formed, but its existence cannot be fully excluded.

␦-Subunit-containing GABA A Receptors
Most of the receptors had a Hill coefficient Յ 1. This cannot be taken as proof for the absence of a second agonist site in these receptors, because depending on the gating mechanism, the Hill coefficient may underestimate this number.

JOURNAL OF BIOLOGICAL CHEMISTRY 7893
did not observe any potentiation of the currents mediated by these receptors.

DISCUSSION
We investigated the architectural role of the ␦-subunit in GABA A receptor pentamers. Preliminary experiments showed that both ␣ 1 -and ␤ 3 -subunits were required to form a functional channel together with the ␦-subunit. Therefore, we focused on the triple subunit combination ␣ 1 ␤ 3 ␦. Functional expression of non-concatenated ␣ 1 ␤ 3 ␦ might theoretically result in multiple subunit arrangements. In this case, subunit concatenation (41-46) is a powerful approach to predefine receptor structure. Two or more subunits can be linked at the DNA level to control subunit composition and arrangement. To construct all possible subunit arrangements and of all studied receptors the pentameric concatenate clearly exceeded our work capacity. As discussed earlier the ␦-subunit is thought to be a ␥-subunit substitute, although it displays the highest degree of homology to the ␤-subunit. Nevertheless, we investigated all variants of the major GABA A receptor isoform ␥␤␣␤␣, where the ␥-subunit, one of the ␣-, or one of the ␤-subunits was replaced by the ␦-subunit.
Assembly of Receptors Following Injection of Individual Dual and Triple Subunit Constructs-Most of the dual and triple subunit constructs when injected alone did not result in current expression, with the exception of concatenated subunits ␤ 3 -␣ 1 and ␤ 3 -␣ 1 -␦ (Fig. 1). It is not clear whether these constructs are able to form tetramers or hexamers, or whether one of the subunits is hanging out, not being incorporated in the pentamer (46). Expression of ␤ 3 -␣ 1 alone resulted in a current that was potentiated less than 2-fold by THDOC. The current mediated by ␤ 3 -␣ 1 -␦/␤ 3 -␣ 1 (R5), which contains this dual subunit construct was potentiated about 7-fold. Also ␤ 3 -␣ 1 -␦ differed from ␤ 3 -␣ 1 -␦/␣ 1 -␤ 3 (R1) in this respect. These observations indicate that in the presence of suitable assembly partners, the mis-formation does not take place. To prove this, we constructed the pentamers ␤ 3 -␣ 1 -␦/␣ 1 -␤ 3 (P1) and ␤ 3 -␣ 1 -␦-␤ 3 -␣ 1 (P5). The functional properties of pentameric receptors were found to be similar as for the respective receptors composed of dual and triple subunit constructs with respect to sensitivity for GABA. In summary, we conclude that the assembly pathway is influenced by the co-expressed subunits.
The situation in the case of ␤ 3 -␦-␤ 3 /␤ 3 -␣ 1 (R7) is less clear. This receptor resulted in currents of slightly larger amplitude as ␤ 3 -␣ 1 . The current showed likewise relatively little stimulation by THDOC. The EC 50 was determined as 40 Ϯ 18 M, and the Hill coefficient as 0.8 Ϯ 0.1. Again these parameters are reminiscent of the current mediated by ␤ 3 -␣ 1 . Therefore, we assume that R7 is probably not formed, but its formation with similar properties as ␤ 3 -␣ 1 cannot be fully excluded.
Abundance of the Different ␦-Subunit-containing Receptors-Our functional study on ␣ 1 ␤ 3 ␦ GABA A receptors should be compared with a structural study on ␣ 4 ␤ 3 ␦ GABA A receptors. Using atomic force microscopy Barrera et al. (40) determined stoichiometry and subunit arrangement of these receptors expressed in tsA 201 cells. They showed that ␣␤␣␦␤ counterclockwise is the predominant subunit arrangement around the pore when viewed from the extracellular space with 21% of the population exhibiting a distinct subunit arrangement of ␣␤␣␤␦. Only a very small number of receptor entities were analyzed, and these numbers should therefore be taken with care. The above study was done at a structural level, whereas we focused on the function of ␦-subunit-containing receptors. If it is assumed that ␣ 1 is similar to ␣ 4 , ␣␤␣␦␤ receptors correspond to ␤ 3 -␣ 1 -␦(/)␤ 3 -␣ 1 (R5/P5) and ␣␤␣␤␦ to ␣ 1 -␤ 3 -␣ 1 /␤ 3 -␦ (R2) in our study. From the present experiments, it is difficult to conclude the relative abundance of the three expressing receptors. Subunit concatenation may affect expression levels. Although the ␤ 3 -␣ 1 -␦-␤ 3 -␣ 1 (P5) receptor with the ␦-subunit in the ␥-subunit position produces the largest current amplitudes, this receptor has an EC 50 for GABA about 12-fold higher than that of non-concatenated ␣ 1 ␤ 3 ␦ receptors. Nevertheless, active, non-concatenated ␣ 1 ␤ 3 ␦ receptors probably constitute a mixture of ␤ 3 -␣ 1 -␦/␣ 1 -␤ 3 (R1), ␣ 1 -␤ 3 -␣ 1 /␤ 3 -␦ (R2), and ␤ 3 -␣ 1 -␦/␤ 3 -␣ 1 (R5), where R2 is only active in the presence of neurosteroids. It should be noted that we cannot fully exclude that in addition ␤ 3 -␦-␤ 3 /␤ 3 -␣ 1 (R7) or other subunit arrangements that were not analyzed could also be formed. Evidence for the expression of multiple receptors has been obtained for another ␦-subunit-containing receptor, namely ␣ 6 ␤ 2 ␦ (50). Taken together, our findings reveal a unique assembly profile for the ␦-subunit that resembles that of the ⑀-subunit (51) with respect to the fact that both subunits can assume multiple positions in a receptor.
Ability of ␦-Subunit to Contribute to the Formation of an Agonist Site-␣ 1 -␤ 3 -␣ 1 /␤ 3 -␦ (R2) had a Hill coefficient greater than 1, hinting at the presence of more than one agonist site. The major isoform of GABA A receptors has two different agonist binding sites located both at the interface of the ␤-and ␣-subunits (43). Assuming that the binding site is formed at the ␤ 3 -␦ and ␤ 3 -␣ 1 interfaces in ␣ 1 -␤ 3 -␣ 1 /␤ 3 -␦, we introduced a homologous point mutation ␤ 2 Y205S (49) into either of the ␤ 3 -subunits to disrupt both agonist binding sites selectively. Our results indicate the existence of two agonist sites involving the ␤ 3 -␦ and ␤ 3 -␣ 1 interfaces in the ␣ 1 -␤ 3 -␣ 1 /␤ 3 -␦ (R2) receptor. Channel opening also occurs when the receptor is occupied with a single agonist molecule, but is promoted more than 30-fold if occupied by two agonists. Thus, the minus side of the ␦-subunit may contribute to an agonist site, but we cannot exclude that the effect of the mutation in the ␤-subunit is allosterically propagated to the plus side of the ␦-subunit (52). Whether or not the ␦-subunits assume the role of the ␣-subunit in the agonist site is not clear from these data. It is however intriguing that the residue ␣ 1 F64 crucial in the agonist site of ␣ 1 ␤ 2 ␥ receptors (52) is conserved in the homologous position in ␦-subunits. Mutation of this residue will clarify the question.
Summary-In summary, we have shown that GABA A receptors containing the ␣ 1 -, ␤ 3 -, and ␦-subunit have a stoichiometry of 2␣ 1 :2␤ 3 :1␦ and that the ␦-subunit exhibits the ability to promiscuously assemble into different ethanol-insensitive subunit arrangements at least in the Xenopus oocytes. Further, we show that at least one of these ␦-subunit-containing receptors remains silent in the absence of neurosteroid. We have also found that the ␦-subunit can contribute to the formation of an agonist site. In the future, it would be interesting to determine how the ␦-subunit assembles in the brain. It is possible that the arrangement of ␦-subunit-containing receptors in brain is controlled in a region-specific manner.