Comparison of γ-Aminobutyric Acid, Type A (GABAA), Receptor αβγ and αβδ Expression Using Flow Cytometry and Electrophysiology

The subunit stoichiometry and arrangement of synaptic αβγ GABAA receptors are generally accepted as 2α:2β:1γ with a β-α-γ-β-α counterclockwise configuration, respectively. Whether extrasynaptic αβδ receptors adopt the analogous β-α-δ-β-α subunit configuration remains controversial. Using flow cytometry, we evaluated expression levels of human recombinant γ2 and δ subunits when co-transfected with α1 and/or β2 subunits in HEK293T cells. Nearly identical patterns of γ2 and δ subunit expression were observed as follows: both required co-transfection with α1 and β2 subunits for maximal expression; both were incorporated into receptors primarily at the expense of β2 subunits; and both yielded similar FRET profiles when probed for subunit adjacency, suggesting similar underlying subunit arrangements. However, because of a slower rate of δ subunit degradation, 10-fold less δ subunit cDNA was required to recapitulate γ2 subunit expression patterns and to eliminate the functional signature of α1β2 receptors. Interestingly, titrating γ2 or δ subunit cDNA levels progressively altered GABA-evoked currents, revealing more than one kinetic profile for both αβγ and αβδ receptors. This raised the possibility of alternative receptor isoforms, a hypothesis confirmed using concatameric constructs for αβγ receptors. Taken together, our results suggest a limited cohort of alternative subunit arrangements in addition to canonical β-α-γ/δ-β-α receptors, including β-α-γ/δ-α-α receptors at lower levels of γ2/δ expression and β-α-γ/δ-α-γ/δ receptors at higher levels of expression. These findings provide important insight into the role of GABAA receptor subunit under- or overexpression in disease states such as genetic epilepsies.

The lack of consensus regarding GABA A receptor subunit stoichiometry and arrangement likely reflects the highly variable methodologies employed to date. For example, although some studies have evaluated receptors composed of "freely assembled" subunits, others have utilized receptors assembled from concatenated subunits (5,14). Concatenation is unquestionably a powerful experimental approach, often representing the only way to definitively test whether a particular subunit arrangement is functional. However, the results represent forced subunit assembly and therefore must not be interpreted in isolation. Concatenation may reveal what subunit stoichiometries and arrangements are theoretically possible, but whether the receptors adopt these configurations naturally is another matter. Moreover, concatenation has known technical limitations, requiring that extensive control experiments be performed to exclude the possibility of "looped out" subunits or linker sequence cleavage (22,23). Expression systems also commonly differ between studies, with some having used humanderived cell lines, but many others having used Xenopus oocytes to boost protein expression. Additional differences in experimental methodology have included the species of subunits employed (e.g. rat versus human) and the identity of partnering subunits (e.g. ␤2 versus ␤3), which could theoretically affect receptor composition.
To improve our understanding of GABA A receptor biogenesis, we compared the surface and total cellular expression profiles of human ␥2 and ␦ subunits permitted to freely assemble with ␣1 and ␤2 subunits in HEK293T cells using a multimodality approach that included flow cytometry, whole cell patch clamp recording (using both freely assembled and concatenated subunits), and traditional biochemistry techniques. By combining these methodologies, we deduced that ␣␤␥ and ␣␤␦ receptors have similar stoichiometries and arrangements. However, 10-fold lower concentrations of ␦ subunit cDNA were required to recapitulate ␥2 subunit expression profiles, reflecting a slower rate of ␦ subunit degradation. Moreover, we found that ␣␤␥ and ␣␤␦ receptor composition depends on relative subunit availability, with alternative subunit arrangements and stoichiometries likely occurring with increasing levels of ␥ or ␦ subunit expression.

Results
GABA A Receptor ␥2L HA and ␦ HA Subunits Had Different Profiles of Surface Expression When Co-transfected with ␣1 and/or ␤2 Subunits at Equimolar cDNA Ratios-To determine the subunit requirements for receptor surface trafficking, we transfected HEK293T cells with all possible combinations of ␣1, ␤2, ␥2L, and ␦ subunit cDNAs (excluding conditions with ␥2L and ␦ subunit co-transfection), labeled subunits with fluorescently conjugated antibodies, and evaluated cell surface fluorescence levels using flow cytometry. Although there has been ongoing debate in the GABA A receptor literature regarding what subunit cDNA ratios should be transfected in recombinant receptor studies (27)(28)(29), we chose to begin with equimolar ratios because this should approximate the relative gene dosage in vivo (␣1, ␤2, ␥2, and ␦ GABA A receptor subunit genes are autosomal and none has been shown to be imprinted). Because no commercially available antibodies against ␥2 or ␦ subunits were found to be suitable for flow cytometry (secondary to excessive nonspecific binding), the HA epitope (YPYDVP-DYA) was inserted near the N termini of ␥2L and ␦ subunits (see under "Experimental Procedures"), and levels of these subunits were detected using a fluorescently conjugated anti-HA antibody.
Of note, there were two primary reasons for using HEK293T cells as opposed to neurons or neuronal cell lines in these experiments. Most importantly, HEK293T cells have minimal if any endogenous expression of GABA A receptor subunits, the presence of which could confound conclusions reached about receptor assembly. Although a few studies have reported low levels of endogenous ␤3 subunit expression in HEK293 cells, this has not been a consistent finding in the literature (30 -32), and it has never been confirmed in our laboratory despite extensive investigation (data not shown). In contrast, neurons are known to express high and variable levels of multiple endogenous GABA A receptor subunits. HEK293T cells are also ideal subjects for flow cytometry experiments, as they are easily harvested and express reproducible levels of subunit protein. In contrast, harvesting of neurons typically results in loss of neuronal processes and, consequently, loss of many postsynaptic GABA A receptors.
GABA A Receptor ␥2L HA and ␦ HA Subunits Had Different Profiles of Total Cellular Expression When Co-transfected with ␣1 and/or ␤2 Subunits at Equimolar cDNA Ratios-The results in the previous section demonstrated that ␣1 and ␤2 subunit surface expression levels were differentially affected by co-transfection with an equimolar ratio of ␥2L HA or ␦ HA subunit cDNA. To explore whether the different patterns of subunit surface expression reflected differential subunit surface trafficking, differential subunit expression, or perhaps a combination of both, total cellular expression levels were assessed by repeating the previously described flow cytometry experiments following cell membrane permeabilization (Fig. 2). Of note, monitoring total expression levels also served as an important control for the flow cytometry assay, as it confirmed that the monoclonal antibodies being used to detect GABA A receptor subunits were highly specific for their respective subunits or epitopes. The ␣1 subunit antibody, for example, was detected intracellularly in all transfection conditions that included ␣1 subunit cDNA but none that lacked ␣1 subunit cDNA ( Fig. 2A). The same was also true for the ␤2 subunit and HA epitope antibodies (Fig. 2, B and C). HEK293T cells were transfected with various combinations of GABA A receptor subunit cDNAs, and surface expression was evaluated using subunit-specific antibodies and flow cytometry. A-C present representative flow cytometry histograms from cells transfected with the indicated combination of subunit cDNAs (left) and incubated with antibodies raised against ␣1 (A), ␤2/3 (B), GABA A receptor subunits or the HA epitope tag (C). The abscissa indicates fluorescence intensity (FI) in arbitrary units plotted on a logarithmic scale, and the ordinate indicates percentage of maximum cell count (% of maximum). Histograms for cells transfected with subunit combinations (dark gray) and cells transfected with blank vector (light gray) are overlaid. D-F present quantifications of fluorescence intensities from cells transfected with the indicated combination of subunit cDNAs and incubated with antibodies raised against ␣1 (D) or ␤2/3 (E) GABA A receptor subunits or the HA epitope tag (F). Mean fluorescence intensities from cells transfected with blank vector alone were subtracted from mean fluorescence intensities of all other expression conditions. All mock-subtracted fluorescence intensities were normalized to the mock-subtracted fluorescence intensity obtained with ␣1␤2␥2L HA subunit co-expression. **, p Ͻ 0.01; ***, p Ͻ 0.001 versus ␣:␤ and † † †, p Ͻ 0.001 versus ␣:␤:␥2L.
By comparing the patterns of surface and total cellular expression, several conclusions could be drawn. First, it appeared that co-expression of ␣1 and ␤2 subunits was both necessary and sufficient for surface trafficking of all tested subunits except for the ␦ HA subunit, which was trafficked to the cell surface independent of subunit combination. Second, in contrast to ␥2L HA subunits, co-transfection of ␦ HA subunits resulted in a marked decrease in both surface and total cellular expression levels of partnering subunits independent of transfection condition, suggestive of a potent dominant negative effect. Although the basis for this dominant negative effect is uncertain based on the results of these experiments alone, it should be noted that ␦ HA subunit co-transfection even decreased total cellular expression of the ␤2 subunit, which could not reach the cell surface when transfected alone. This suggests that the dominant negative effects of the ␦ HA subunit were, at least in part, mediated by promoting subunit degradation. Third, given that surface and total cellular expression levels of the ␦ HA subunit were essentially constant across transfection conditions (Figs. 1 and 2) and not significantly different compared with expression levels of the ␦ HA subunit transfected alone (despite universally decreased expression levels of partnering subunits), it seems likely that the majority of ␦ HA sub-FIGURE 2. GABA A receptor ␣1, ␤2, ␥2L HA , and ␦ HA subunit total cellular expression was highly sensitive to the presence and identity of partnering subunits. HEK293T cells were transfected with various combinations of GABA A receptor subunit cDNAs, and total cellular subunit expression was evaluated after cell membrane permeabilization using flow cytometry. A-C, fluorescence intensities were quantified from cells transfected with the indicated combination of subunit cDNAs and incubated with antibodies raised against ␣1 (A) or ␤2/3 (B) GABA A receptor subunits or the HA epitope tag (C). Mean fluorescence intensities from cells transfected with blank vector alone were subtracted from mean fluorescence intensities of all other expression conditions. All mock-subtracted fluorescence intensities were normalized to the mock-subtracted fluorescence intensity obtained with ␣1␤2␥2L HA subunit co-expression. , ***, p Ͻ 0.001 versus ␣:␤ and † † †, p Ͻ 0.001 versus ␣:␤:␥2L.
units trafficked to the cell surface were unassembled with cotransfected subunits.
Decreasing the Amount of ␦ HA Subunit cDNA Co-transfected with ␣1 and ␤2 Subunits by 10-Fold Recapitulated ␥2L HA Subunit Surface Expression Patterns-Up to this point, the results of surface and total cellular expression studies raised the possibility that ␥2L and ␦ subunits were incorporated into ternary GABA A receptors quite differently. However, one important similarity could not be ignored; addition of both ␥2L and ␦ subunits had disparate effects on ␣1 and ␤2 surface and total expression levels. Despite the profound dominant negative effect of adding the ␦ subunit on partnering subunit expression, ␤2 subunit expression levels were consistently more reduced than ␣1 levels, a phenomenon also observed for the ␥2L subunit. We therefore hypothesized that differences observed between ␥2L and ␦ subunits simply reflected differences in subunit "availability," as opposed to intrinsic differences in receptor stoichiometry and/or arrangement. Indeed, the relatively constant patterns of surface and total cellular expression observed with ␦ subunit co-transfection (independent of subunit combination) could be a manifestation of marked protein overexpression, which can overload normal cellular trafficking machinery, activate protein degradation pathways, and trigger apoptosis. Consistent with the latter, we noted considerably higher rates of cell death in all conditions involving 1 g of ␦ subunit transfection, but not ␥2L transfection, unless the latter was transfected in considerable excess (10 g; data not shown).
To explore this possibility, surface and total cellular expression levels were again evaluated with flow cytometry but with variable amounts of ␥2L HA or ␦ HA cDNA transfected (Fig. 3). Specifically, the amount of ␣1 and ␤2 subunit cDNA transfected was held constant (1 g per subunit), although the amount of ␥2L HA or ␦ HA cDNA was systematically increased over several orders of magnitude. Of note, evaluation of subunit expression was not possible when more than 1 g of ␦ HA cDNA was transfected, as this resulted in widespread cell death. The results demonstrated that increasing the amount of ␥2L HA or ␦ HA subunit cDNA transfected yielded similar overall patterns of subunit expression but with far less ␦ HA subunit cDNA being required to produce comparable expression levels (i.e. the ␦ HA subunit cDNA titration curves were all left-shifted compared with those of the ␥2L HA subunit). For instance, ␣1 subunit surface levels remained relatively stable when low levels of ␥2L HA subunit cDNA were transfected (Fig. 3A) and began progressively decreasing only after Ͼ0.3 g of ␥2L HA subunit cDNA was transfected (Fig. 3A, black line). A similar "plateau" phase was seen with increasing ␦ HA subunit cDNA levels; however, the subsequent progressive decrease began when the ␦ HA subunit cDNA level was only Ն0.03 g (Fig. 3A, gray line). Surface expression levels of the ␤2 subunit also responded similarly to increasing amounts of ␥2L HA or ␦ HA cDNA transfected, but again with far less ␦ HA subunit cDNA required to produce comparable expression levels. The pattern, however, was distinct from that seen with the ␣1 subunit. Instead of an early plateau phase, increasing amounts of ␥2L HA or ␦ HA subunit cDNA caused steady concentration-dependent decreases in ␤2 subunit surface levels (Fig. 3B), suggestive of ␥2L HA or ␦ HA subunit incorporation into ternary subunits occurring primarily at the expense of ␤2 subunits. Finally, ␥2L HA and ␦ HA subunit surface levels also had similar patterns with increasing cDNA levels, but with the ␦ HA curve again appearing left-shifted compared with the ␥2L HA curve. For both subunits, surface levels increased over a range of cDNA levels, peaked, and then decreased (Fig.  3C). However, peak subunit surface expression occurred with only 0.03 g of ␦ HA cDNA, as compared with 1 g of ␥2L HA cDNA. Notably, peak ␥2L HA and ␦ HA subunit surface levels were not achieved until after ␣1 subunit surface levels began to decline, suggesting that at higher transfection ratios, incorporation of ␥2L HA and ␦ HA subunits into ternary receptors could also occur at the expense of ␣1 subunits. Of note, although the highest transfection ratios for the ␥2L HA subunit (1:1:3 and 1:1:10) included higher amounts of ␥2L HA subunit cDNA than other transfection conditions (5 and 12 g, respectively; see "Experimental Procedures"), the overall patterns of ␣1, ␤2, and ␥2L HA subunit surface and total cellular expression were similar to those obtained at the highest levels of ␦ HA subunit expression, where total cDNA levels were maintained at 3 g.
Total cellular subunit expression patterns over a range of ␥2L HA and ␦ HA subunit cDNA levels were generally similar to surface expression patterns, noting that total cellular subunit protein levels decreased to a lesser extent at the highest amounts of ␥2L HA or ␦ HA cDNA. Levels of ␣1 subunits began declining when more than 0.3 g of ␥2L HA subunit cDNA or 0.01 g of ␦ HA subunit cDNA was transfected (Fig. 3D), whereas levels of ␤2 subunits began declining when more than 0.01 g of either ␥2L HA or ␦ HA subunit cDNA was transfected (Fig. 3E). Interestingly, the total cellular expression patterns of ␥2L HA and ␦ HA subunits were somewhat different from their surface expression patterns (Fig. 3F). Here, ␥2L HA subunit levels peaked when 3 g (as opposed to 1 g) of cDNA was transfected, after which they declined from that peak by ϳ25% with 10 g of cDNA transfected (as opposed to the 80% decrease in surface expression). Similarly, although ␦ HA subunit surface levels peaked when 0.03 g of cDNA was transfected and declined by about 75% when 1 g of cDNA was transfected, ␦ HA total cellular levels steadily increased over the entire range of cDNA amounts.
These findings supported several conclusions. First and foremost, the observation that increasing amounts of ␥2L HA and ␦ HA subunit cDNA had nearly identical effects on the expression profiles of partnering subunits argues against significant differences in how these subunits were incorporated into ternary receptors. Both ␥2L HA and ␦ HA subunits appeared on the cell surface at the expense of ␤2 subunits, noting a range of ␥2L HA /␦ HA subunit cDNA amounts over which ␣1 subunit levels did not change, ␤2 subunit levels decreased, and ␥2L HA /␦ HA subunit levels increased. At higher transfection ratios, both ␥2L HA and ␦ HA subunit surface expression levels continued to rise sharply despite concomitant decreases in both ␣1 and ␤2 subunit levels, suggesting that incorporation into ternary receptors could also occur at the expense of ␣1 subunits when sufficient competing ␥2L HA or ␦ HA subunit was available. In addition, both ␥2L HA and ␦ HA subunits appeared to require co-transfection with ␣1 and ␤2 subunits for maximal surface expression, as surface levels were considerably lower at all titra-tion points when ␥2L HA and ␦ HA subunits were transfected alone (data not shown). Finally, both ␥2L HA and ␦ HA subunits appeared capable of exerting a profound dominant negative effect on co-transfected subunit levels at the highest transfection ratios, as surface levels of all subunits declined after ␥2L HA / ␦ HA subunit levels peaked, after which there was also considerable cell death not seen in cells treated with equivalent levels of transfection reagent and blank pcDNA vector (data not shown).

Fluorescence Resonance Energy Transfer (FRET) Analysis Suggested Similar Patterns of Subunit Adjacency in ␣1␤2␥2L
and ␣1␤2␦ Containing Receptors-FRET is an established methodology for monitoring protein-protein interactions (35). In contrast to conventional biochemical techniques (e.g. co-immunoprecipitation), FRET can be used to monitor proteins in their native conformations and to identify direct protein interactions. Although FRET can be measured by spectrofluorimetry and microscopy, flow cytometry offers several advantages over these techniques. Unlike spectrofluorimetry, measurements can be performed on individual cells, and importantly, donor emission can easily be distinguished from sensitized emission of the acceptor. Although microscopy allows the subcellular localization of protein interactions to be evaluated, the technique is less sensitive, analysis is labor-intensive and poorly quantitative, and selecting regions of interest is highly subjective. Flow cytometry, in contrast, allows for rapid, quantitative, and unbiased analysis of FRET in large cell popu-   . GABA A receptor ␣1, ␤2, and ␥2L HA or ␦ HA subunits had similar surface and total expression levels and patterns but required markedly different amounts of ␥2L HA or ␦ HA cDNA. Flow cytometry was used to evaluate surface and total expression of GABA A receptor subunits in HEK293T cells transfected with ␣1, ␤2, and varying amounts of ␥2L HA or ␦ HA subunit cDNAs. A-C, surface expression levels of ␣1 (A), ␤2 (B), and ␥2L HA (C) subunits were evaluated in cells transfected with 1 g of ␣1 subunit cDNA, 1 g of ␤2 subunit cDNA, and 0.01-10 g of ␥2L HA subunit cDNA. Mean fluorescence intensities from cells transfected with blank vector alone were subtracted from mean fluorescence intensities of all other expression conditions. All mock-subtracted fluorescence intensities were normalized to the mock-subtracted fluorescence intensity obtained with 1 g of ␣1, 1 g of ␤2, 1 g of ␥2L HA subunit co-expression. D-F, total expression levels of ␣1 (D), ␤2 (E), and ␥2L HA (F) subunits were evaluated in cells transfected with 1 g of ␣1 subunit cDNA, 1 g of ␤2 subunit cDNA, and 0.001-1 g of ␦ HA subunit cDNA. Mean fluorescence intensities from cells transfected with blank vector alone were subtracted from mean fluorescence intensities of all other expression conditions. All mock-subtracted fluorescence intensities were normalized to the mock-subtracted fluorescence intensity obtained with 1 g of ␣1, 1 g of ␤2, 0.1 g of ␦ HA subunit co-expression.
lations and permits the simultaneous analysis of other cellular properties (e.g. viability) (36). Based on homology modeling to nAChRs, GABA A receptors are thought to assemble into pseudo-symmetrical pentamers (37). As a result, each subunit is predicted to have two "adjacent" subunits and two "non-adjacent" subunits. Given the crystal structure of the homomeric ␤3 subunit GABA A receptor (38), the N termini (where our subunit-and epitope-specific antibodies bind) of adjacent subunits in ternary ␣1␤2␥2L and ␣1␤2␦ receptors are likely separated by ϳ30 -40 Å, whereas those of non-adjacent subunits are likely separated by ϳ50 -60 Å (accounting for slight differences in N-terminal length between subunit subtypes and due to epitope tagging). Because FRET efficiency is inversely proportional to the sixth power of distance (35), adjacent subunits would be expected to yield a more robust FRET signal as compared with non-adjacent subunits. For ternary receptors, there are six possible subunit combinations that can be probed with a FRET assay (e.g. for ␣1␤2␥2L receptors, FRET could potentially occur between ␣1 subunits, between ␤2 subunits, between ␥2L subunits, between ␣1 and ␤2 subunits, between ␣1 and ␥2L subunits, and between ␤2 and ␥2L subunits). By analyzing the FRET signals between all possible subunit combinations, a FRET "profile" can be generated for each isoform. If ␣␤␥ and ␣␤␦ receptors have different subunit stoichiometries and/or arrangements, then distinct FRET profiles would be predicted.
To perform the FRET assay, transfected cells were co-stained with equimolar mixtures of donor-conjugated (Alexa-555) and acceptor-conjugated (Alexa-647) subunit-specific antibodies (Förster radius for this fluorophore pair was 51 Å). The resulting FRET signal was evaluated using flow cytometry, as described previously (39). Specifically, the FRET signal represented emission detected from the acceptor fluorophore following excitation of the donor fluorophore, using an excitation wavelength that could not significantly excite the acceptor fluorophore directly. Of note, the exact distance between donor and acceptor fluorophores in this experimental paradigm was uncertain, as the configuration of antibodies bound to the GABA A receptor subunits and the location of fluorophores conjugated to the antibodies were unknown. This uncertainty precluded a priori estimation of FRET efficiency between these fluorophores.
To determine whether there was a significant contribution of FRET from non-adjacent subunits in this experimental paradigm, concatenated ␣1-␤2 subunits were co-transfected with the ␤2 HA subunit, forcing the ␤2 HA -␣1-␤2-␣1-␤2 subunit arrangement. Co-staining this receptor with donor-conjugated anti-HA and acceptor-conjugated anti-␣1 antibodies, FRET was detected between the ␤ HA subunit and the concatenated ␣1 subunit (n ϭ 4), confirming FRET between adjacent subunits. In contrast, co-staining with a mixture of donor-and acceptorconjugated anti-␣1 antibodies demonstrated no FRET, indicating that FRET could not occur between non-adjacent subunits. FRET was also not observed when receptors were co-stained with a mixture of donor-and acceptor-conjugated anti-HA antibodies, indicating that FRET could not occur between pentamers (data not shown). However, we cannot exclude the possibility that antibodies bound to freely assembled subunits have slightly different orientations compared with those bound to concatameric subunits, which might permit FRET between non-adjacent subunits. With that said, the primary goal of these experiments was to compare the FRET profiles of ␣␤␥ and ␣␤␦ receptors, in the hope of exposing differences in their underlying stoichiometries and/or arrangements, and not to explicitly to deduce them.
10-Fold Difference in ␥2L HA and ␦ HA Subunit Expression Levels Did Not Reflect Differences in the Efficiency of Subunit Incorporation into Ternary Receptors-There are two possible explanations for the lower amounts of ␦ subunit cDNA required to recapitulate the flow cytometry expression patterns seen with the ␥2L subunit (Figs. 3 and 4). Either ␦ subunits are incorporated more efficiently into ternary GABA A receptors than ␥2L subunits or higher ␦ subunit levels are available for incorporation per unit of cDNA transfected, possibly because of increased subunit transcription, increased subunit translation, or decreased subunit degradation.
To test the first of these hypotheses, HEK293T cells were transfected with varying concentrations of ␥2L HA or ␦ HA subunit cDNA in the absence of ␣1 or ␤2 subunits, and total cellular expression levels were evaluated using flow cytometry. Indeed, if the previous findings were related to more efficient incorporation of ␦ subunits into ternary receptors, then the discrepancy between ␥2L and ␦ subunit levels should not be observed when the subunits are expressed alone. However, we found that total cellular expression of ␦ HA subunits (solid gray line) was significantly higher than that of ␥2L HA subunits (solid black line) at all points when less than 1 g of subunit cDNA was transfected (Fig. 5B, circles; n ϭ 4). For instance, fluorescence levels detected when 0.03 g of ␦ HA cDNA was transfected were nearly identical to those detected when 0.3 g of ␥2L HA cDNA was transfected. Thus, the 10-fold difference in ␥2L HA and ␦ HA subunit expression levels (Fig. 3) persisted in the absence of partnering subunits, suggesting that differences in the efficiency of subunit incorporation did not account for the previous findings.
10-Fold Difference in GABA A Receptor ␥2L HA and ␦ HA Subunit Total Cellular Expression Could Not Be Accounted for by Differences in cDNA Transcription or Translation-To determine whether the difference between ␥2L HA and ␦ HA expres-sion levels was a result of more efficient transcription, real time PCR was performed on cells transfected over the same range of subunit cDNA used in the single subunit studies. Transcript levels were determined by normalized difference in cycle number fold increase. As cDNA levels increased, mRNA levels for ␥2L HA and ␦ HA subunits increased similarly and proportionally ( Fig. 5A; n ϭ 4), indicating that equivalent amounts of ␥2L HA and ␦ HA cDNA did not produce different amounts of protein because of differences in transcription efficiency.
To determine whether the difference between ␥2L HA and ␦ HA expression levels was a result of more efficient translation, we began by closely comparing the sequences of the cDNA constructs employed. Indeed, levels of protein expression in heterologous systems are known to be exquisitely sensitive to the design of the nucleic acid construct. For instance, if the ␥2L HA coding sequence and untranslated regions were significantly longer than those of the ␦ HA subunit, then equimolar amounts of plasmid DNA might not represent equimolar amounts of subunit cDNA. Full sequencing confirmed that the ␥2L HA and ␦ HA subunit inserts (translated and untranslated sequences) were approximately the same length. However, the sequences of the immediate 5Ј-untranslated regions differed slightly between ␥2L HA and ␦ HA subunit cDNAs. This could be problematic, because the 3 bp preceding and 2 bp following a start codon constitute the Kozak sequence, which contributes to the efficiency of translation initiation. Specifically, ribosome binding is strongly enhanced by the presence of purines at the Ϫ3 and ϩ4 positions with respect to the start codon (40). In our cDNA constructs, the Kozak sequence of the ␥2L HA subunit cDNA was TCC(AUG)A, whereas the corresponding sequence of the ␦ HA subunit cDNA was GCC(AUG)G; consequently, the ␦ HA subunit would be predicted to be translated more efficiently than the ␥2L HA subunit.
To determine whether the Kozak sequence differences could account for the findings, the Kozak sequences of the ␥2L and ␦ subunits were swapped. Therefore, the plasmids were engineered such that the ␥2L HA construct contained the Kozak sequence GCC(AUG)G and the ␦ HA construct contained the Kozak sequence TCC(AUG)G (␥2L(T-3G) HA and ␦(G-3T) HA , respectively. The single subunit titration experiments were then repeated using the ␥2L(T-3G) HA and ␦(G-3T) HA constructs. Surprisingly, the Kozak sequence mutations had little effect on total subunit expression levels. There was no significant difference between ␥2L HA (Fig. 5B, solid black line, circles) and ␥2L(T-3G) HA (Fig. 5B, dotted black line, triangles; n ϭ 4) or between ␦ HA (Fig. 5B, solid gray line, circles; n ϭ 4) and ␦(G-3T) HA (dotted gray line, triangles; n ϭ 4) subunit levels at any tested amount of subunit cDNA. Therefore, the 10-fold difference in GABA A receptor ␥2L HA and ␦ HA subunit expression could not be attributed to differences in subunit synthesis at the stage of transcription or translation initiation.
However, it is possible that the subsequent rate of protein synthesis differed significantly and thus was responsible for the higher levels of ␦ subunit protein. To explore this possibility, HEK293T cells transfected with 1 g of either ␥2L HA or ␦ HA subunit cDNA were incubated for 0 -20 min in media containing 150 Ci/ml [ 35 S]methionine ( Fig. 6A; n ϭ 4). At 5-min intervals, radiolabeled GABA A receptor subunit protein was precipitated by incubation with anti-HA beads. After elution, protein was subjected to SDS-PAGE and exposed to a phosphor screen. Integrated band density for each time point was calculated and normalized to the integrated band density of ␥2L HA subunits that were radiolabeled for 20 min (Fig. 6B; n ϭ 4). When subunit expression levels at each time point were directly compared, ␦ HA subunit levels were greater than ␥2L HA subunit levels at the 5-, 10-, and 15-min time points (p Ͻ 0.01) but not at the 20-min time point. This difference was surprising given that both subunits were engineered with identical plasmids and therefore regulated by identical promoters. Although difficult to explain, it should be noted that observed differences between ␥2L HA and ␦ HA subunit levels in these experiments were relatively small and were insufficient to account for the 10-fold difference in expression levels seen in prior experiments. Moreover, when the synthesis curves were fitted using a mixed procedure model produced in consultation with the Vanderbilt University Department of Biostatistics, the synthesis curve slopes (which are most indicative of the rate of subunit synthesis) were not significantly different (p ϭ 0.099).  . GABA A receptor ␥2L HA and ␦ HA subunits were synthesized at similar rates, and when newly synthesized, were degraded at similar rates. Metabolic labeling was used to assess the synthesis and degradation rates of ␥2L HA and ␦ HA subunits. A and B, HEK293T cells expressing equivalent amounts of ␥2L HA or ␦ HA subunits were cultured for 0 -20 min in media containing [ 35 S]methionine. Subunit protein was isolated from cell lysates by immunoprecipitation and separated by SDS-PAGE. A presents a representative gel exposure; B presents a quantification of band intensity (IDV) averaged from four separate experiments. Band intensities were normalized to those of the 20-min incubation condition. C, HEK293T cells expressing equivalent amounts of ␥2L HA or ␦ HA subunits were cultured for 20 min in media containing [ 35 S]methionine. To assess degradation rates of this newly synthesized protein population, radioactive media were subsequently replaced by regular media, and cells were returned to incubators for 0 -6 h. Subunit protein was isolated from cell lysates by immunoprecipitation and separated by SDS-PAGE. C presents a representative gel exposure, and D presents a quantification of band intensity (IDV) averaged from four separate experiments. Band intensities are normalized to that of the 0-min chase condition. tion, we next explored whether differences in subunit degradation could account for the findings using a pulse-chase assay. HEK293T cells transfected with 1 g of either ␥2L HA or ␦ HA subunit cDNA were incubated for 20 min in radioactive media. Subsequently, the radioactive medium was replaced with normal culture medium, and cells were returned to the incubator. After 0 and 1-4 or 6 h, ␥2L HA and ␦ HA subunit proteins were extracted, immunoprecipitated, and processed as described above (Fig. 6C; n ϭ 4). Integrated band density for each subunit was calculated and normalized to the 0-h time point. Both subunits had a half-life of ϳ1.5 h and decayed with essentially identical time courses (Fig. 6D; n ϭ 4). It should be noted that a similar decay course has been reported for ␥2S subunits (41). Thus, it initially seemed that subunit degradation could not account for the differing levels of ␥2L HA and ␦ HA subunit expression.

10-Fold Difference in GABA
However, the pulse-chase technique only measures the degradation rate of proteins synthesized during the labeling period (i.e. subunits that may not have had sufficient time to be completely processed and trafficked). To evaluate degradation rate of the entire cellular pool of ␥2L HA and ␦ HA subunits, a variation of the previously described flow cytometry assay was employed. HEK293T cells were transfected with ␥2L HA and ␦ HA subunit cDNA and cultured as in previous experiments. Two days after transfection, 100 g/ml cycloheximide (CHX) 3 was added to the culture medium to inhibit protein synthesis for 0 -6 h before being harvested. Cells were then permeabilized, incubated with antibodies, and analyzed using flow cytometry. In contrast to the results obtained using the pulsechase protocol, the flow cytometry CHX assay demonstrated that ␥2L HA subunits degraded more rapidly than ␦ HA subunits ( Fig. 7; n ϭ 4). During the 1st h of treatment, both subunits declined similarly. After 1 h, ␥2L HA subunit levels (Fig. 7, solid black line, circles) had decreased to 77.2 Ϯ 4.4% of 0-h levels, whereas ␦ HA subunit levels (solid gray line, circles) had decreased to 81.9 Ϯ 4.8% of 0-h levels. After this point, degradation time courses diverged. After 6 h, ␦ HA subunit levels remained stable at ϳ80% of 0-h levels (6 h of CHX, 86.0 Ϯ 7.1%). In contrast, after 3 h of CHX treatment, ␥2L HA subunit levels were approximately half (53.8 Ϯ 2.3%) of 0-h levels, and they remained similar through the rest of the 6-h treatment period (6 h, 49.1 Ϯ 3.5% of 0-h levels).
One possible explanation for these findings relates to the different cellular distributions of ␥2L HA and ␦ HA subunits when transfected alone (Fig. 1). Specifically, in the absence of ␣1 and ␤2 subunits, ␥2L HA subunits were mostly retained intracellularly, but many ␦ HA subunits were trafficked to the cell surface. Given that a substantial fraction of the intracellular subunit pool is likely destined for proteasomal degradation, the different degradation rates of ␥2L HA and ␦ HA subunits might simply reflect their different cellular distributions. If so, ␥2L HA and ␦ HA subunits would be expected to degrade at similar rates when expressed together with ␣1 and ␤2 subunits, which enabled maximal surface expression of all subunits. However, repeating the experiment in the setting of ␣1 and ␤2 subunit co-expression did not affect degradation rates of either ␥2L HA or ␦ HA subunits ( Fig. 7; n ϭ 4). The ␥2L HA subunit population (Fig. 7, dashed black  Very Low Levels of the ␦ Subunit Were Required to Eliminate the Functional Signature of ␣1␤2 Receptors-The findings in the previous sections indicated that very low levels of ␦ subunit cDNA were necessary to achieve high levels of expression, reflecting slower degradation as compared with the ␥2L subunit. This suggested that low levels of the ␦ subunit would be required to eliminate the functional signature of ␣1␤2 receptors. To test this hypothesis, HEK293T cells were co-transfected with fixed levels of ␣1 and ␤2 subunit cDNA (1 g per subunit) and variable concentrations (0.01-1 g) of ␦ subunit cDNA. All cDNA constructs included non-HA-tagged sequences to allow comparison with previously reported results, as well as subsequent studies using the available concatenated cDNA constructs. GABA was applied for 4 s (Fig. 8A), and whole cell currents were analyzed for peak current amplitude (Fig. 8B) as well as macroscopic kinetic properties, including rise time (Fig. 8C), extent of desensitization (Fig. 8D), and deactivation time course (Fig. 8E). It should be noted that although experiments were conducted with the electrophysiologist blinded to transfection conditions, this proved impossible for cells transfected with the highest tested levels of ␦ subunit cDNA (1 g) due to widespread cell death and abnormal morphology. Furthermore, the effects of Ͼ1 g of ␦ subunit 3   cDNA could not be tested due to nearly universal death and poor membrane integrity of surviving cells.
Although flow cytometry using cells transfected with ␦ subunit cDNA showed surface ␦ subunit expression, these cells had no detectable GABA-evoked currents (n ϭ 5). Similarly, no GABA-evoked current was obtained from cells co-transfected with ␣1 and ␦ subunit cDNAs (n ϭ 6) or from cells co-transfected with ␤2 and ␦ subunit cDNAs (n ϭ 4) (data not shown). Along with the flow cytometry results reported above, these data suggest a negligible contribution of any endogenous ␤ subunit expression to our results. Very low levels of ␦ subunit cDNA produced significant changes in macroscopic current properties when co-transfected with ␣1 and ␤2 subunit cDNAs (Fig. 8). Cells expressing only ␣1 and ␤2 subunits produced currents with peak amplitudes of 2811 Ϯ 921 pA (n ϭ 7). However, when just 0.01 g of ␦ subunit cDNA was included, peak current amplitudes increased significantly to 6099 Ϯ 880 pA (n ϭ 6, p Ͻ 0.01 compared with 1:1:0 g condition). When 0.03 g of ␦ subunit cDNA was included, peak current amplitude was 2477 Ϯ 453 pA (n ϭ 8), similar to the 1:1:0 g condition. When still more ␦ subunit cDNA was included, peak current amplitudes continued to decline, with equimolar transfection yielding peak current amplitudes of only 68.8 Ϯ 24.1 pA (n ϭ 6, p Ͻ 0.01 compared with 1:1:0 g condition). Although this precipitous drop in current amplitude mirrors the loss of receptor surface expression seen in the flow cytometry experiments, the possibility that altered receptor kinetic properties were contributing to loss of current amplitude cannot entirely be excluded.
It is commonly accepted that ␣1␤ x receptor currents desensitize far more extensively than ␣1␤ x ␦ receptor currents (1). In agreement, increasing levels of ␦ subunit cDNA resulted in progressively decreasing extents of current desensitization. In the 1:1:0 g transfection condition, currents desensitized by 80.5 Ϯ 3.2% (n ϭ 7), and in the 1:1:1 g transfection condition, currents desensitized by only 9.2 Ϯ 3.0% (n ϭ 6). In all conditions other than 1:1:0.01 g, desensitization was significantly different from that of the 1:1:0 g condition (p Ͻ 0.001). Finally, the time course of current deactivation was fitted for all transfection conditions and weighted deactivation time constants were calculated. In previous studies, ␣1␤3 and ␣1␤3␦ receptors deactivated at similar rates (1), so it was perhaps unsurprising that there were no significant differences in deactivation time constants among all transfection conditions. However, there was a trend toward slower deactivation with increasing ␦ subunit cDNA, which was maximal when the 1:1: 0.03 ratio of ␦ subunit cDNA was transfected (217.5 Ϯ 64.2 ms, n ϭ 8, compared with 1:1:0, 104 Ϯ 6.8 ms; n ϭ 6), after which there was a change in the trend toward faster deactivation with the highest level of ␦ subunit cDNA.
Taken together, these findings suggested that co-transfection of only 0.03 g of ␦ subunit cDNA was sufficient to essentially eliminate the functional signature of ␣1␤2 receptors. Whether this represented a truly homogeneous ␣1␤2␦ receptor population, however, remained unclear, as macroscopic current kinetic properties continued to change beyond this point, raising the possibility that alternative ␣␤␦ receptor isoforms could be formed in the presence of high ␦ subunit cDNA levels. Indeed, transfection with the highest amounts of ␦ subunit cDNA resulted in alterations of macroscopic current kinetic properties. Unfortunately, the marked associated decrease in current amplitude with increasing levels of ␦ subunit cDNA significantly limited detailed kinetic analysis, particularly at the highest transfection levels.
Low Levels of the ␥2 Subunit Were Required to Eliminate the Functional Signature of ␣1␤2 Receptors-To determine the cDNA transfection ratio at which the functional signature of ␣1␤2 receptors was eliminated in the presence of the ␥2 subunit, and to explore the possibility that the kinetic properties of ␥2 subunit-containing receptors were also sensitive to transfection ratio, HEK293T cells were co-transfected with ␣1 and ␤2 subunits (1 g per subunit) and variable amounts (0.01-10 g) of the ␥2S subunit. (Of note, using the ␥2S instead of the ␥2L subunit allowed us to directly compare the electrophysiology results to those obtained from the available tandem constructs used in subsequent experiments. No significant difference was found in the kinetic properties of these splice variants; data not shown.) GABA was applied for 4 s (Fig. 9A), and whole cell currents were analyzed for peak current amplitude (Fig. 9B) as well as macroscopic kinetic properties including rise time (Fig.  9C), extent of desensitization (Fig. 9D), and deactivation time course (Fig. 9E). As with the ␦ subunit experiments, the electrophysiologist was blinded to transfection conditions, but again, this proved impossible for cells transfected with the highest tested levels of ␥2S subunit cDNA (10 g) due to widespread cell death.
As with the ␦ subunit electrophysiology experiments, peak current amplitudes were significantly affected by even low amounts of ␥2S subunit cDNA transfected. For example, although cells transfected with only ␣1 and ␤2 subunit cDNAs had peak current amplitudes of 814 Ϯ 266 pA (n ϭ 14) (Fig. 9, A  and B), addition of just 0.01 g of ␥2 subunit cDNA significantly increased peak current amplitude to 3510 Ϯ 682 pA (n ϭ 17, p Ͻ 0.05). Although this marked increase in current amplitude with very low levels of the ␥2 subunit appeared surprising given the minimal associated change in receptor surface levels, this likely reflects the much higher charge transfer mediated by ␣1␤␥2 receptors (ϳ7-fold higher than ␣1␤2 receptors) secondary to increased single channel conductance and gating differences (29). For example, if just 15% of surface ␣1␤2 receptors are converted into ␣1␤2␥2L receptors, current amplitude would be expected to nearly double. Higher ␥2 subunit cDNA levels yielded similar increases in current amplitudes (all ␥2 subunit cDNA amounts from 0.01 to 3 g produced currents that were significantly larger than ␣1␤2 currents), with the largest peak current observed in the 1:1:0.3 g transfection condition (5866 Ϯ 761 pA; n ϭ 14, p Ͻ 0.001 compared with ␣1␤2). Interestingly, there was a strong trend toward decreasing amplitude with high ␥2 subunit cDNA amounts above 0.3 g. For the 1:1:10 g transfection condition, peak current amplitude was only 2870 Ϯ 480 pA (n ϭ 21), which was ϳ50% lower than the peak current amplitude seen in the 1:1:0.3 g transfection condition (4530 Ϯ 483 pA; n ϭ 25). Although it seems likely that this decrease in current amplitude was at least partly accounted for by loss of receptors on the cell surface given the flow cytometry results, the degree to which current amplitude was reduced is less than that seen for surface expression, raising the possibility that intrinsic channel kinetic properties had changed at the highest transfection ratios.
According to most reports, ␣␤ and ␣␤␥ receptors both desensitize extensively; however, ␣␤ receptors desensitize more rapidly. To determine whether a shift from ␣1␤2 to ␣1␤2␥2 receptor populations could be detected by changes in desensitization kinetics, 1 mM GABA was applied for 4 s to transfected cells, and the desensitization time course of resulting currents was fitted with up to four exponential components. The percent of all desensitization contributed by the two shorter components (1 and 2, Յ16 and 17-125 ms, respectively) was summed and defined as fast desensitization (Fig.  9D). For ␣1␤2 receptors, 65% of all desensitization was contributed by the two shorter components, and this fraction dropped significantly when 0.01 g of ␥2 subunit cDNA was co-expressed (50 Ϯ 5%, p Ͻ 0.05). Only 0.03 g of ␥2 subunit cDNA was necessary to reduce fast desensitization to levels statisti-cally indistinguishable from those produced by 1 g of ␥2 subunit cDNA (32 Ϯ 5 and 23 Ϯ 5%, respectively). Interestingly, 10 g of ␥2 subunit cDNA further reduced fast desensitization, to a level that was significantly lower than 1 g (5 Ϯ 1%, p Ͻ 0.01). Thus, similar to the results for current rise time, analysis of fast desensitization demonstrated that low levels of ␥2 subunit cDNA were sufficient to eliminate the functional signature of ␣1␤2 receptors and that increasing levels of ␥2 subunit cDNA continued to change kinetic properties again, implying the presence of one or more alternative receptor isoforms when ␥2 subunit cDNA was transfected in excess.
The weighted time constant of deactivation also changed dramatically in response to the amount of ␥2 subunit cDNA that was transfected (Fig. 9E). Specifically, when more ␥2 sub-   (42), whereas ␣␤␥ receptor currents are not. In contrast, ␣␤ receptors are insensitive to diazepam (DZP), reflecting the absence of an ␣-␥ subunit interface, whereas diazepam significantly enhances ␣1,2,3,5␤␥ receptor GABA-evoked currents (43,44). (Of note, ␣␤␦ receptors are only partially Zn 2ϩ -sensitive and DZP-insensitive, so these techniques could not be used to differentiate ␣␤ and ␣␤␦ receptors.) To determine how much ␥2 subunit cDNA was necessary to eliminate the pharmacological signature of ␣␤ receptors, peak current amplitude in response to GABA (1 mM, 4 s) was recorded (I max (GABA)); Zn 2ϩ (10 M) was pre-applied for 10 s, and peak current amplitude was recorded again although GABA and Zn 2ϩ were co-applied (I max (GABA ϩ Zn 2ϩ )). Zn 2ϩ inhibition was quantified by dividing I max (GABA ϩ Zn 2ϩ ) by I max (GABA) (Fig. 9F). As expected, cells transfected with only ␣1 and ␤2 subunits produced currents that were inhibited strongly by Zn 2ϩ co-application (peak current amplitude was 19 Ϯ 1% of those evoked by GABA alone). When 0.01 or 0.03 g of ␥2 subunit cDNA was included, peak current amplitude was partially Zn 2ϩ -sensitive (32 Ϯ 13 and 84 Ϯ 9% of I max (GABA), respectively; p Ͻ 0.001 compared with ␣␤). When Ն0.1 g of ␥2 subunit cDNA was included, peak current amplitude was maximally Zn 2ϩ -insensitive, again suggesting that low ␥2 subunit levels were sufficient to eliminate the ␣␤ receptor population.
As mentioned previously, ␣␤␥ receptors are inhibited by Zn 2ϩ but enhanced by DZP. We first characterized the GABA concentration-response relationships for a subset of stoichiometries. The GABA EC 50  To determine how much ␥2 subunit cDNA was necessary to produce a DZP-sensitive receptor population, the percent enhancement of ϳEC 20 GABA-evoked peak current amplitude by 1 M DZP was evaluated (Fig. 9G). Even 0.01 g of ␥2 subunit cDNA permitted substantial DZP potentiation of peak current amplitude (134 Ϯ 13% of control current), and 0.03 g was sufficient to produce potentiation (204 Ϯ 18%) indistinguishable from 1 g (260 Ϯ 26%) or 10 g (257 Ϯ 56%).
These experiments indicated that low levels of ␥2 subunit cDNA were necessary to eliminate the pharmacological signa-tures of ␣1␤2 receptors. However, it should be emphasized that the observed transition point to a predominantly ␣␤␥ receptor population may be exaggerated in these experiments due to the aforementioned higher charge transfer associated with ␣1␤2␥2 receptors (see above for ␣1␤2␥2 electrophysiology experiments). Nevertheless, the results clearly demonstrate that excess ␥2 subunit was not required to eliminate the pharmacological signature of ␣1␤2 receptors. Interestingly, DZP potentiation reached a plateau at a transfection ratio of ϳ1:1:0.3 g, meaning that any alternative ␣␤␥ receptor populations introduced at higher transfection ratios (as suggested by the aforementioned changes in macroscopic current kinetics) must be similarly DZP-sensitive, implying that the ␣-␥ subunit interface must be preserved in those isoforms.
Expression of Concatenated Subunit Constructs Demonstrated That Alternate ␣1␤2␥2 GABA A Receptor Stoichiometries Were Functional-The flow cytometry and electrophysiology subunit titration studies suggested the presence of alternative receptor isoforms at high levels of ␥2 or ␦ subunit expression, with kinetic properties distinct from those seen with low levels of ␥2 or ␦ subunit expression. Importantly, these conclusions were based entirely on results from experiments where receptor subunits were permitted to assembly freely. However, to confirm our suspicion that alternative functional stoichiometries and subunit arrangements were possible, we turned to cDNA constructs encoding concatameric GABA A receptor subunits, acknowledging the known technical limitations of this approach. For example, although concatenated subunits have the distinct advantage of constraining the arrangement of GABA A receptor subunit proteins, concern has been raised that linking peptides could potentially be cleaved, releasing individual subunits and invalidating any stoichiometric constraints. There is also concern that concatamers could remain physically intact but only partially incorporate into a receptor (i.e. become looped out of the receptor), again resulting in lack of stoichiometric constraint.
To address the first of these concerns, tandem constructs were expressed in HEK293T cells and detected with immunoblotting (of note, the flow cytometry assay could not be used to detect concatenated subunits, as the linkers markedly decreased subunit antibody binding, which is targeted to the N terminus). Proteins were identified at ϳ50-, 100-, and 150-kDa bands when individual GABA A receptor subunits, double-subunit concatamers, or triple-subunit concatamers were individually expressed, respectively, indicating that linker cleavage had not occurred. Moreover, no abnormal lower molecular weight bands were consistently identified to suggest partial protein cleavage (data not shown). With that said, it should be noted that despite adjusting cDNA transfection levels to compensate for construct size (see "Experimental Procedures"), individual subunits were expressed at much lower levels when they were incorporated into concatamers than when they were expressed as individual subunits. Consequently, we did not directly compare amplitudes of currents recorded from cells expressing concatemeric subunits to those expressing freely assembled subunits. Based on the presence or absence of currents, however, electrophysiological recording should expose which sub-unit combinations can theoretically form functional receptors and which cannot.
Currents were recorded from HEK293T cells expressing various combinations of ␤2-␣1-␤2, ␤2-␣1-␥2, and ␣1-␥2 concatamers together with ␣1, ␤2, and/or ␥2 monomers (see under "Experimental Procedures" for details of tandem construction). The combinations chosen were designed to constrain receptor assembly such that some conditions should not produce pentameric receptors, some should only produce "canonical" GABA A receptor isoforms (␤-␣-␥-␤-␣), and some could produce only alternative isoforms (of particular interest were those containing multiple ␥2 subunits). For each condition, the possible isoforms and kinetic properties of resulting whole cell currents were listed in Fig. 10. Of note, we focused on ␥2 subunitcontaining concatameric constructs for several reasons. First, they had already been generated, optimized, and validated by the Sigel laboratory in combination with ␤2 subunits by the time of data acquisition, allowing for direct comparison to our results from freely assembled subunits, whereas ␦ subunit concatameric constructs were novel and only available in combination with ␤3 subunits, which could have confounded results due known differences in trafficking and assembly characteristics. Second, although ␣1␤x␥2 receptors are the most abundant combination found in brain, the ␣1␤2␦ subunit combination has a limited distribution. Most ␦ subunit-containing receptors are partnered with ␣4 or ␣6 subunits, but using a non-␣1 subunit subtype would have confounded our ability to elucidate the rules for ␦/␥2 subunit incorporation. Third, given that expression of concatenated subunits tends to be low, ␥2 subunit-containing concatameric constructs were preferred over ␦ subunit concatameric constructs due to their higher charge transfer, which should offset the lower expression levels. Finally, it should be emphasized that the flow cytometry data demonstrated nearly identical expression patterns with ␥2 and ␦ subunits when expressed at equivalent levels, suggesting that ternary ␣1␤2␥2 and ␣1␤2␦ receptors have similar subunit stoichiometries and arrangements.
To address the technical concern that individual subunits within a concatamer may "loop out," allowing double subunit constructs to incorporate into receptors as single subunits (or triple subunit constructs to incorporate as double subunits), we first evaluated currents from cells transfected with one tandem construct alone (e.g. ␤-␣-␤, ␣-␥, and ␤-␣-␥; Fig. 10, rows 1, 5,  and 8, respectively). In all conditions, no GABA-evoked currents were recorded. We next evaluated a concatamer combination that should only produce the canonical ␤-␣-␥-␤-␣ iso-  (I max ), rise time (Rise), desensitization (% Fast Desens), and effect of DZP on deactivation ( deact prolongation)). Note that peak current amplitudes should be interpreted with caution, as they are likely heavily influenced by receptor surface expression, which is markedly reduced for concatemeric constructs. Only possible subunit arrangements that included a properly oriented ␤-␣ interface (for GABA binding) and ␣-␥ interface (for benzodiazepine binding) are illustrated. Potential arrangements containing ␤-␤ and/or ␥-␥ subunit interfaces were deemed unlikely based on the FRET results.
form (␣-␥ ϩ ␤-␣-␤; Fig. 10, row 7), and we confirmed that it yielded a robust current (1012 Ϯ 272 pA; n ϭ 8). These control experiments suggested that these concatameric constructs assembled as predicted within HEK293T cells and could therefore be used to explore the possibility of alternative subunit stoichiometries and arrangements.

Discussion
Flow Cytometry Provided an Efficient Quantitative Method for Evaluating GABA A Receptor Subunit Expression-Among ion channels, GABA A receptors are remarkable for their complexity. The 19 subunit subtypes, many of which are co-expressed in individual neurons, can theoretically produce a myriad of heteropentameric receptor isoforms. Determining which GABA A receptor subunit combinations can traffic to the cell surface, and what the stoichiometries and arrangements are of these receptors, has thus presented a fascinating yet frustrating problem for investigators. Here, we demonstrate the utility of a flow cytometry assay for a high-throughput quantitative evaluation of subunit expression resulting from numerous transfec-tion combinations. Flow cytometry can also serve as a surrogate for some traditional biochemistry techniques, as illustrated in our experiments comparing subunit degradation rates.
Using these flow cytometry-based assays, we demonstrated that ␥2L and ␦ subunits were incorporated into ternary GABA A receptors similarly when expressed at equivalent protein levels as follows: both required co-expression with ␣1 and ␤2 subunits for maximal surface expression; both incorporated into ternary receptors preferentially at the expense of ␤2 subunits; both yielded identical subunit FRET profiles, suggesting similar underlying stoichiometries and arrangements; and both exerted potent dominant negative effects on co-transfected subunits when expressed at high levels. The most striking difference between ␥2L and ␦ subunits was their stability, the latter having a slower rate of degradation, and consequently higher levels of surface and total cellular expression at any given amount of cDNA transfected. We propose that this could represent an important regulatory mechanism for extrasynaptic ␣␤␦ receptors, which mediate "tonic" inhibitory currents and might therefore benefit from increased stability. Interestingly, ␦ subunits were also capable of reaching the cell surface when transfected alone, a phenomenon previously only described for ␤1/3, ␥2S, and 1-3 subunits (33,(45)(46)(47)(48).
kinetic and pharmacological properties. Although there has been some work with concatameric constructs confirming that double-␦ subunit receptors are theoretically possible (11,13), direct comparison with our results is limited, as the identity of partnering subunits differed in those studies (e.g. ␣1 versus ␣4; ␤2 versus ␤3). Of note, differences in partnering subunit identity may also explain why prior studies have reported conflicting ␣␤␦ stoichiometries with increasing transfection ratios, some supporting the canonical 2:2:1 ratio (8,9) but others supporting ratios up to 1:1:3 (11).
This study provides important insights into how the relative expression levels of individual subunit genes help determine the subunit stoichiometry and arrangement of functional surface GABA A receptors. Although we found a limited cohort of alternative pentameric assemblies, there are undoubtedly further constraints imposed when actually expressed in neurons. An important future direction of this work will be determining which, if any, of the alternative ␣␤␥ and ␣␤␦ isoforms exist in vivo. Very early immunoprecipitation studies of ␥ subunits from rodent brain suggested the presence of receptors containing multiple ␥2 subunits (16). However, the presence of multiple ␥2 subunits is predicted in our model to result in loss of the GABA-binding site and gain of a benzodiazepine-binding site (Fig. 11, right column), in contrast to early work with receptors isolated from bovine brain, which demonstrated a 2:1 ratio of GABA to benzodiazepine-binding sites (50). The presence of double-␦ subunit receptors in vivo has not yet been explored to our knowledge. If alternative isoforms do occur outside of heterologous expression systems, it will be interesting to determine their relative proportions, subcellular localization, and functional and pharmacological properties and to determine how these properties are affected by disease-causing mutations (e.g. under/overexpressing GABA receptor subunit mutations associated with epilepsy). It will also be of interest to explore the mechanistic bases for the altered subunit stoichiometries and arrangements from the standpoint of receptor assembly.
Low Levels of ␥2L and ␦ Subunit cDNA Were Sufficient to Eliminate the Functional Signature of ␣1␤2 Receptors-There is ongoing debate in the GABA A receptor literature regarding the optimal subunit transfection ratios necessary to achieve a homogeneous population of ternary ␣␤␥ or ␣␤␦ receptors. This is of particular importance for investigations aimed at characterizing receptor kinetics and pharmacology, as heterogeneous receptor populations confound analysis. Because ␣␤ receptors are expressed quite efficiently, some groups consider it necessary to transfect the third (e.g. ␥ or ␦) subunit in excess, assuming this will ensure a homogeneous ternary receptor population (27,28). In contrast, other groups have found that equimolar transfection ratios of ␥/␦ subunit cDNA are sufficient to eliminate the functional signature of ␣1␤2 receptors (1,29).
Our results demonstrate that ␥2L or ␦ subunit overexpression is unnecessary and likely counterproductive given the aforementioned effects on partnering subunit expression and cell viability. For the ␥2 subunit, the flow cytometry and electrophysiology data collectively suggested that the ␣1␤2 receptor population was eliminated early in the cDNA titrations, likely by an ␣:␤:␥ subunit cDNA transfection ratio of 1:1:0.3. Of note, many of the kinetic and pharmacological properties of ␣1␤2 receptors were eliminated at even lower ␥2 subunit transfection ratios, likely reflecting 7-fold higher charge transfer of ␣1␤2␥2 receptors that obscures the functional signature of ␣1␤2 receptors. For the ␦ subunit, the results were more dramatic, with the ␣1␤2 receptor population appearing eliminated by ␣:␤:␦ subunit cDNA transfection ratio of only 1:1:0.03, reflecting the much higher stability of the ␦ subunit. It should be emphasized, however, that eliminating the ␣1␤2 receptor population does not guarantee formation of homogeneous receptor populations. The results of the flow cytometry experiments suggest that receptor heterogeneity is already present with ␣1 and ␤2 subunit co-expression (Fig. 11, left column). Addition of ␥2 or ␦ subunits likely converts this to yet another heterogeneous receptor population (Fig. 11, middle column). In other words, receptor heterogeneity may be more of the rule rather than the exception. Although these findings could be artifactual related to heterologous expression, work with under-and overexpressing GABA A receptor subunit epilepsy-associated mutations further support the idea that native receptor combinations may include a continuum of ␣1 2 ␤x 3 -␣1 2 ␤x 2 ␥2 1 -␣1 2 ␤2 1 ␥2 2 stoichiometries (51).
Interestingly, the electrophysiology titration experiments appear to have reconciled a long-standing debate in the GABA A receptor literature. Although we have consistently observed ␣␤␥ receptor currents that desensitize extensively (52), others have reported poorly desensitizing ␣␤␥ receptor currents (53). Notably, different experimental conditions were employed in these experiments, with highly desensitizing currents having been obtained with equimolar transfection ratios, and poorly desensitizing currents having been obtained in the setting of ␥2 subunit overexpression. The results from this study indicate that ␣␤␥ receptors can have different functional properties depending on the amount of ␥2 subunit cDNA transfected, reflecting alterations in receptor stoichiometry. Low levels of ␥2 subunit expression yield single ␥2 subunit-containing receptors that undergo extensive desensitization, whereas high levels of ␥2 subunit expression yield double-␥2 subunit-containing receptors that yield less desensitizing currents (Figs. 9 and 10). Prior results thus likely appeared contradictory because currents were effectively being recorded from different receptor populations.
Although not explicitly verified using concatameric constructs, our electrophysiology results suggest that ␣␤␦ receptors also have altered kinetic properties at higher transfection ratios (Fig. 8). Given the similarities between the expression patterns of ␣␤␥ and ␣␤␦ receptors, the functional changes observed presumably also reflect altered ␣␤␦ receptor subunit composition, with receptors formed at the highest transfection ratios likely containing two ␦ subunits.

Experimental Procedures
Cell Culture and Expression of Recombinant GABA A Receptors-Human GABA A receptor ␣1 (NM_000806), ␤2 (NM_000813), ␥2S (NM_000816), ␥2L (NM_198904), and ␦ (NM_000815) subunits were individually subcloned into the pcDNA3.1ϩ mammalian expression vector (Invitrogen). Because of the lack of a highly specific, commercially available antibody targeting an extracellular domain on the ␥2 and ␦ subunits, the HA (YPYDVPDYA) epitope was inserted between amino acids 5 and 6 of the mature ␦ subunit and between amino acids 4 and 5 of the ␣1, ␤2, and ␥2 subunits. This resulted in N-terminal sequences of MNDIGYPYDVPDYADYVGS and QKSDYPYDVPDYADDYED for the mature ␦ and ␥2 subunits, respectively, based on nucleotide sequencing and SignalP predictions. The predicted signal-mature peptide cleavage site was not altered by the HA sequence insertion (data not shown). This insertion site was selected for its known minimal effect on receptor expression and function (24). Although we did not directly evaluate the effect of inserting the HA epitope on ␥2 and ␦ subunit stability, we inferred that stability of these tagged constructs was similar to those of non-tagged subunits given the lack of significant effect on the expression profiles of co-expressed subunits. Specifically, a subset of experiments performed in Figs. 1-3 using HA-tagged ␥2 and ␦ subunits was performed in parallel using non-tagged ␥2 and ␦ subunits, and the profiles of ␣1 and ␤2 surface and total cellular expression were found to be nearly identical (data not shown). HA-tagged ␣1 and ␤2 constructs were generated for the purposes of FRET experiment in a similar manner, with the HA epitope inserted between amino acids 4 and 5 of the mature peptide. This resulted in an N-terminal sequence of QPSLYPYDV-PDYAQDELKDNTTVFT and QSVNYPYDVPDYADPSNM-SLVKE for the mature ␣1 and ␤2 subunits, respectively. As with HA-tagged ␥2 and ␦ subunits, HA tagging of the ␣1 and ␤2 did not appear to affect expression of partnering subunits (data not shown). Concatenated subunit plasmids were a generous gift from Prof. Erwin Sigel. Synthesis of these constructs has been described previously (6). Briefly, the plasmids contained cDNA from rat ␣1, ␤2, and ␥2S subunits connected by 10 -26 amino acid linkers containing glutamine, alanine, and proline residues. The tandem construct in which the C terminus of the ␤2 subunit was linked to the N terminus of the ␣1 subunit will be denoted "␤-␣" and so forth. The coding region of each vector was sequenced by the Vanderbilt University Medical Center DNA Sequencing Facility and verified against published sequences.
HEK293T cells (American Type Culture Collection, Manassas, VA) were maintained at 37°C in humidified 5% CO 2 , 95% air using Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum (Invitrogen), 100 IU/ml penicillin (Invitrogen), and 100 g/ml streptomycin (Invitrogen). Cells were plated at a density of ϳ10 6 cells per 10-cm culture dish (Corning Glassworks, Corning, NY) and passaged every 2-4 days using trypsin/EDTA (Invitrogen). For flow cytometry and electrophysiology experiments, cells were plated at a density of 4 ϫ 10 5 cells per 6-cm culture dish (Corning Glassworks) and transfected ϳ24 h later with equal amounts (1 g/subunit) of subunit cDNA using FuGENE 6 (Roche Diagnostics) per manufacturer's protocol. In conditions where less than 3 g of subunit cDNA was transfected, empty pcDNA3.1 vector was added such that a total of 3 g of cDNA was used for each experimental condition (thus, the "mock" transfection condition consisted of 3 g of empty pcDNA 3.1 vector cDNA). In the subset of experiments using ␣:␤:␥ cDNA transfection ratios of 1:1:3 and 1:1:10, an additional 2 and 9 g of ␥2 subunit cDNA was transfected, bringing the total amount of cDNA transfected to 5 and 12 g, respectively. Tandem construct transfection levels were adjusted for the size of the plasmid to deliver equimolar amounts of each subunit cDNA. Specifically, for each 1 g of single-subunit cDNA transfected, 1.4 g of triple-subunit and 1.2 g of double-subunit concatamer cDNA was transfected. Finally, an additional 1 g of pHook-1 cDNA (encoding the cell surface antibody sFv) was included for electrophysiology experiments so that positively transfected cells could be selected ϳ24 h later by immunomagnetic bead separation, as described previously (25). Following selection, cells were re-plated at low density on 35-mm dishes for electrophysiological recording the following day.
Electrophysiology-Whole cell patch clamp recordings were obtained at room temperature from lifted cells that were maintained during recordings in a bath solution consisting of (in mM) the following: 142 NaCl, 8 KCl, 6 MgCl 2 , 1 CaCl 2 , 10 glucose, and 10 HEPES (pH adjusted to 7.4; 325-330 mosM). All chemicals used for solution preparation were purchased from Sigma. Recording pipettes were pulled from thin-walled borosilicate capillary glass (Fisher) on a Sutter P-2000 micropipette electrode puller (Sutter Instruments, San Rafael, CA) and firepolished with a microforge (Narishige, East Meadow, NY). When filled with a pipette solution consisting of (in mM) 153 KCl, 1 MgCl 2 , 5 EGTA, 10 HEPES, and 2 MgATP (pH 7.3; and osmolarity 300 -310 mosM) and submerged in the bath solution, yielding open tip resistances of ϳ1.5-2 megohms and a chloride equilibrium potential (E Cl ) of ϳ0 mV. Currents were recorded at a holding potential of Ϫ20 mV using an Axopatch 200B amplifier (Molecular Devices, Foster City, CA), low-pass filtered at 2 kHz using a 4-pole Bessel filter, digitized at 10 kHz using the Digidata 1322A (Molecular Devices), and stored off line for analysis. Because ␣␤␦ receptor currents tended to be smaller in amplitude, those recordings (Fig. 8) were obtained at a holding potential of Ϫ50 mV. GABA was prepared as a stock solution. Working solutions were made on the day of the experiment by diluting stock solutions with the bath solution.
Analysis of Current Kinetic Properties-For those cells with small (Ͻ50 pA) currents, rise time, desensitization, and deactivation were not determined. Analysis of very small currents (usually Ͻ10 pA) were determined to be significantly different from zero using a one-sample t test. To minimize series resistance error, we recorded current amplitudes from all cells, but we only analyzed the kinetic properties of currents recorded from cells with peak currents less than 6 nA. This cutoff allowed us to minimize the holding voltage error to 1-2 mV (26). Current amplitudes and 10 -90% rise times were measured using the Clampfit 9 software package (Molecular Devices). The desensitization and deactivation time courses of GABA A receptor currents were fit using the Levenberg-Marquardt least squares method with up to four component exponential functions of the form shown in Equation 1, ane ͑Ϫt/n͒ ϩ C (Eq. 1) where t is time; n is the best number of exponential components; a n is the relative amplitude of the nth component; n is the time constant of the nth component; and C is the residual current at the end of the GABA application. The individual time constants fell naturally into four distinct groups (1 Ͻ 16 ms, 2 17-125 ms, 3 126 -800 ms, and 4 Ͼ 800 ms). The first two time constants ( 1 and 2 ) were considered to define "fast desensitization." Additional components were accepted only if they significantly improved the fit, as determined by an F-test automatically performed by the analysis software on the sum of squared residuals. The time course of deactivation was summarized as a weighted time constant, defined by Equation 2, ann an (Eq. 2) Solution exchange time was defined as the time for an opentip liquid-junction current to increase from 10 to 90% of its maximum value. Data were reported as mean Ϯ S.E. One-way analysis of variance followed by a Dunnett's multiple comparison test was used to compare results to the 1:1:0 and 1:1:1 g transfection conditions, as indicated.
Flow Cytometry-Cells were harvested ϳ48 h after transfection using 37°C trypsin/EDTA (Invitrogen) and placed immediately in 4°C FACS buffer composed of PBS (Mediatech), 2% fetal bovine serum (FBS) (Invitrogen), and 0.05% sodium azide (VWR Scientific). Cells were then transferred to 96-well plates, where they were washed twice in FACS buffer (i.e. pelleted by centrifugation at 450 ϫ g, vortexed, and resuspended). For surface protein staining, cells were incubated in antibody-containing FACS buffer for 1 h at 4°C, washed in FACS buffer three times, and resuspended in 2% w/v paraformaldehyde (Electron Microscopy Sciences). For total protein staining, samples were first fixed and permeabilized using Cytofix/Cytoperm (BD Biosciences) for 15 min. After washing twice with Permwash (BD Biosciences) to remove residual fixative, cells were resuspended in antibody-containing Permwash for 1 h at 4°C. Following incubation with antibody, samples were washed four times with Permwash and twice with FACS buffer before resuspension in 2% paraformaldehyde. The anti-␣1 antibody was obtained from Millipore (clone bd24), conjugated to the Alexa-647 fluorophore using an Invitrogen kit, and used at 4 g/ml for surface subunit staining and 2 g/ml for total subunit staining. The anti-␤2 antibody was obtained from Millipore (clone 62-3G1) and used at 8 g/ml for surface subunit staining and 4 g/ml for total subunit staining. Because anti-␤2 antibody conjugation proved inefficient, an anti-IgG1-Alexa-647 secondary antibody was used at a 1:500 dilution for most experiments. The anti-HA antibody (clone 16B12) was obtained from Covance as an Alexa-647 conjugate and used at a 1:250 dilution for surface subunit staining and a 1:500 dilution for total subunit staining.
Samples were run on an LSR II flow cytometer (BD Biosciences). For each staining condition, 50,000 cells were analyzed. Nonviable cells were excluded from analysis based on forwardand side-scatter profiles, as determined from staining with 7amino-actinomycin D (Invitrogen). The Alexa-555 fluorophore was excited using a 535-nm laser and detected with a 575/26 bandpass filter. The Alexa-647 fluorophore was excited using a 635-nm laser and detected with a 675/20 bandpass filter. Data were acquired using FACSDiva (BD Biosciences) and analyzed off-line using FlowJo 7.1 (Treestar). To compare surface and total expression levels of GABA A receptor subunits, the mean fluorescence intensity of mock-transfected cells was subtracted from the mean fluorescence intensity of each positively transfected condition. The remaining fluorescence was then normalized to that of a control condition, yielding a relative fluorescence intensity ("Relative FI"). Statistical significance was determined using a one-sample t test using a hypothetical mean of 1 (because data in each condition were normalized to wildtype expression). Data were expressed as mean Ϯ S.E.
For protein degradation experiments, cells were plated at a density of 2 ϫ 10 5 cells per 3-cm culture dish and transfected as described above, but with a total of 1 g of cDNA for each experimental condition. Approximately 48 h after transfection, 100 l of 0.1% cycloheximide (Sigma) was added to culture dishes, which were subsequently returned to the 37°C incubator for the times indicated in the figure legends. After incubation, cells were harvested, permeabilized, stained, and subjected to flow cytometry as described previously.
Radiolabeling, Immunoprecipitation, and SDS-PAGE-HEK293T cells were plated and transfected with ␥2L HA or ␦ HA subunit cDNA as described above. Two days after transfection, the culture medium was replaced with methionine-free medium for 30 min and then replaced with medium containing 150 Ci/ml [ 35 S]methionine, and cells were returned to the incubator. For synthesis studies, plates were removed after 5, 10, 15, or 20 min, immediately placed on ice, and washed with both non-radioactive media and PBS. Membranes were lysed using radioimmunoprecipitation assay buffer (RIPA buffer: 50 mM Tris-HCl, pH 7.4, 1% Triton X-100, 250 mM NaCl, 5 mM EDTA) containing protease inhibitor mixture (Sigma), and insoluble components were removed by centrifugation at 15,000 ϫ g for 20 min. The ␥2L HA and ␦ HA subunits were incubated overnight with an anti-HA affinity gel (Sigma) and eluted using 125 g/ml anti-HA peptide (Sigma). Proteins were separated using SDS-PAGE (10% BisTris gel). The dried gel was exposed to a phosphor screen for 2 days and imaged using a Typhoon phosphorimager (Molecular Dynamics/GE Healthcare). The bands then were quantified using ImageJ. Degradation studies were performed identically except that after addition of radioactive medium the cells were returned to 37°C for 1-4 or 6 h.
Statistical Analysis-All statistical analyses were performed using GraphPad Prism (version 5.02, La Jolla CA) software. Unless otherwise specified, data are expressed as mean Ϯ S.E., and one-way analysis of variance with Tukey's post hoc test was used to determine whether there were significant differences (p Ͻ 0.05) among different transfection conditions.