A GABAA Receptor α1 Subunit Tagged with Green Fluorescent Protein Requires a β Subunit for Functional Surface Expression*

γ-Aminobutyric acid, type A (GABAA) receptors, the major inhibitory neurotransmitter receptors in the central nervous system, are heteropentameric proteins assembled from distinct subunit classes with multiple subtypes, α(1–6), β(1–4), γ(1–3), δ(1), and ε(1). To examine the process of receptor assembly and targeting, we tagged the carboxyl terminus of the GABAA receptor α1 subunit with red-shifted enhanced green fluorescent protein (EGFP). Xenopus oocytes were injected with cRNA of this fusion protein, α1-EGFP, alone or in combination with cRNA of GABAA receptor β2, γ2, or β2+γ2 subunits. Within 72 h after injection, EGFP fluorescence was visible in all fusion protein-injected cells. The fluorescence was associated with the plasmalemma only when the β2 subunit was co-injected with α1-EGFP. Texas Red-conjugated immunolabeling of EGFP on nonpermeabilized cells demonstrated that EGFP was localized extracellularly. Hence, the COOH terminus of the α1subunit is extracellular. Two-electrode voltage clamp of α1-EGFPβ2- and α1-EGFPβ2γ2-injected oocytes demonstrates that these cells express functional receptors, with EC50 values for GABA and diazepam similar to wild-type receptors. Thus, a COOH-terminal tag of the α1 subunit appears to be functionally silent, providing a useful marker for studies of GABAA receptor expression, assembly, transport, targeting, and clustering. Moreover, the β2 subunit is required for receptor assembly and surface expression.

GABA A 1 receptors are the major inhibitory neurotransmitter receptors in the mammalian brain and are members of a ligand-gated ion channel superfamily (1), which includes receptors for acetylcholine, glycine, and serotonin. The native GABA A receptor is likely to be a heteropentameric protein (2) assembled from several different classes and isoforms of GABA A receptor subunits, including 6␣, 4␤, 3␥, 1␦ and 1⑀ subunit subtypes (3)(4)(5)(6). Hydropathy plots of the sequences of all the subunits predict an extracellular NH 2 -terminal domain of about 200 amino acids, four putative membrane-spanning domains, referred to as M1, M2, M3, and M4, a cytoplasmic domain of variable length between M3 and M4, and an extra-cellular carboxyl terminus. Theoretically, a large number of different pentameric subunit combinations derived from the various subunit subtypes can be present in the brain. Localization of the GABA A receptor gene products, including mRNA by in situ hybridization and polypeptides by immunocytochemistry, has revealed that the various subunits and subunit subtypes show different regional as well as developmental distributions (7)(8)(9)(10)(11) with many neuronal cell types expressing multiple receptor subunits (12,13). GABA A receptors consisting of different subunit isoforms are most likely responsible for the diversity of inhibitory synaptic responses observed in some regions of the brain, and it is possible that different types of GABA A receptors are present even within single neurons (14). Thus, elucidating the cellular and molecular mechanisms involved in the regulation of the assembly, membrane targeting and subcellular distribution of GABA A receptors is important for ultimately understanding how neurons control their fast inhibitory responses.
In general, antibody labeling of receptors has been used to study receptor trafficking and localization. This approach, however, can be labor intensive and is generally not useful for monitoring receptors in intracellular compartments. More recently, green fluorescent protein (GFP) from the jellyfish Aequorea victoria has been used in a variety of studies to examine gene expression and protein subcellular localization (15)(16)(17). GFP is particularly useful because it fluoresces in living cells without requiring additional reagents or cofactors, and its fluorescence is resistant to photobleaching. GFP has been used successfully as a protein tag on both the amino and carboxyl termini of a wide range of cytosolic and membrane-bound proteins (for review, see Cubitt et al. (18)) and provides a way to directly measure levels of protein expression, to visually identify cells expressing the protein, and to localize the protein as it is being processed.
To specifically examine GABA A receptor expression, assembly and membrane targeting, we attached red-shifted GFP (EGFP, CLONTECH) to the COOH terminus of the GABA A receptor ␣ 1 subunit. This fusion protein, ␣ 1 -EGFP, was expressed in Xenopus oocytes alone or with ␤ 2, ␥ 2 , or ␤ 2 ϩ␥ 2 GABA A receptor subunit combinations. Oocytes expressing the fusion protein are clearly fluorescent and show distinct patterns of fluorescence, depending on the subunits coinjected. Voltage clamp studies provide evidence that the fusion protein forms a functional channel when coexpressed with ␤ 2 or ␤ 2 ␥ 2 subunits, and that the GFP tag is functionally silent. In addition, GFP immunolabeling on nonpermeabilized cells provides direct experimental evidence that the COOH terminus of the ␣ 1 subunit is extracellular, consistent with the four-transmembrane subunit topology model based on hydrophobicity analysis.

MATERIALS AND METHODS
Construction of ␣ 1 -EGFP Fusion cDNA-By recombinant polymerase chain reaction, the stop codon of ␣ 1 was changed to proline, which created a new AgeI site, using the oligonucleotide 5Ј-CTAAAAGAAC-CGGTTGATGGGGTGTGGGGGC-3Ј. A polymerase chain reaction product using this primer and a 5Ј-upstream ␣ 1 -specific primer resulted in a 581-base pair fragment of ␣ 1 , with a modified COOH terminus. The ␣ 1 fragment was digested with AgeI and BamHI and subcloned into the pEGFP-N1 vector (CLONTECH) which fused the EGFP gene, in frame, to the COOH terminus of the ␣ 1 fragment. This recombinant plasmid was digested with BamHI and XbaI, and the resulting ␣ 1 -EGFP fragment was subcloned into a pGH19 vector, which contained the entire wild-type ␣ 1 cDNA (19) for expression in Xenopus oocytes. Thus, the 3Ј end of the wild-type ␣ 1 subunit, from the BamHI site to the stop codon, was replaced with the ␣ 1 -EGFP fusion fragment. The ␣ 1 -EGFP fusion cDNA was verified by restriction digest and double-stranded DNA sequencing using standard techniques (20).
Voltage-Clamp Analysis-Oocytes under two electrode voltage-clamp (V hold ϭ Ϫ80 mV) were perfused continuously with ND96 recording solution at a rate of 5 ml/min. Drugs and reagents were dissolved in ND96. The stock diazepam solution was made in Me 2 SO. GABA responses were scaled for run-down or run-up by comparison to a low, nondesensitizing concentration of drug applied just prior to the drug concentration tested. Diazepam potentiation was recorded at approximately EC 7 to EC 20 for GABA. Potentiation is defined as (I (GABAϩDZ) / I (GABA) Ϫ 1), where I (GABAϩDZ) is the current response in the presence of diazepam (DZ) and I (GABA) is the control GABA current. Standard two-electrode voltage-clamp recording was performed using a Ge-neClamp 500 (Axon Instruments) interfaced to a computer with an IT-16 A/D device (Instrutech). Electrodes were filled with 3 M KCl and had a resistance of 0.5-1.5 megohms.
Data acquisition and analysis were performed using AxoData, Axo-Graph (Axon Instruments), and Prism software (Graphpad). Dose-response data were fit to the following four parameter equation derived from the Hill equation: Y ϭ Min ϩ (Max Ϫ Min)/(1 ϩ 10 (log EC 50 Ϫ X)⅐n H ) ), where Max is the maximal response, Min is the response at the lowest drug concentration tested, X is the logarithm of drug concentration, EC 50 is the half-maximal response, and n H is the Hill coefficient.
Immunolabeling of Oocytes-Oocytes were immunoreacted either live or following fixation in 10% neutral-buffered formalin. Removal of the vitelline membrane was not found to be necessary. The cells were blocked in 1% bovine serum albumin in ND96 or 1% bovine serum albumin in phosphate-buffered saline (2.7 mM KCl, 1.5 mM KH 2 PO 4 , 0.5 mM MgCl 2 , 137 mM NaCl, 14 mM Na 2 HPO 4 , pH 7.1). Antibodies were diluted in the corresponding blocking buffer. The primary antibody was an anti-GFP antibody (CLONTECH), diluted at 1:2000; the secondary antibody was a biotinylated goat anti-rabbit (Jackson), diluted 1:1000. The final incubation was in Texas Red-conjugated streptavidin (Jackson), diluted at 1:400. After the last washes, all oocytes were maintained in ND96.
Fluorescence Analysis and Confocal Laser-scanning Microscopy-Oocytes were mounted under a glass coverslip on a glass dual-well slide.
FIG. 1. Visualization of GABA A receptor ␣ 1 subunit distribution in Xenopus oocytes using an ␣ 1 -EGFP fusion protein. Injection of oocytes with the cRNA of EGFP or the fusion protein ␣ 1 -EGFP, alone or in combination with the cRNA of the ␤ 2 , ␥ 2 , or both ␤ 2 ϩ ␥ 2 subunits produces cells with strong green fluorescence. In oocytes expressing EGFP, ␣ 1 -EGFP, or ␣ 1 -EGFP␥ 2 , the fluorescent signal is distributed throughout the cell (upper row). In oocytes expressing ␣ 1 -EGFP␤ 2 or ␣ 1 -EGFP␤ 2 ␥ 2 , the green fluorescent signal appears to be localized to the surface plasma membrane (bottom row).
The slide was fastened upside down on the stage of a Bio-Rad MRC 1024 confocal laser scanning microsope and visualized with a 20ϫ lens. EGFP excitation/emission was achieved with a filter set (488 nm/510 nm) designed for fluorescein detection. Texas Red excitation/emission was achieved with a filter set specific for Texas Red (595 nm/615 nm). All images were recorded at the same adjustments of laser power and photomultiplier sensitivity and were later processed by using Adobe Photoshope software (ADOBE Systems, Mountain View, CA) with identical values for contrast and brightness.

RESULTS
Distinct Cellular Distribution of the ␣ 1 -EGFP Subunit-To determine whether the ␣ 1 -EGFP fusion cRNA produced a fluorescent protein, oocytes were injected with six different groups of cRNA, ␣ 1 -EGFP alone, ␣ 1 -EGFP ϩ ␤ 2 , ␣ 1 -EGFP ϩ ␥ 2 , ␣ 1 -EGFP ϩ ␤ 2 ϩ ␥ 2 , ␤ 2 alone, and EGFP alone (Fig. 1). Two to four days after injection, the oocytes were visualized by epifluorescence microscopy for the development of green fluorescence. Oocytes injected with ␤ 2 alone show very low levels of autofluorescence and defined base-line fluorescence. Oocytes injected with EGFP alone were used as positive controls and produced fluorescent protein that appears to be distributed throughout the cytoplasm of the cells. 2 Oocytes injected with the ␣ 1 -EGFP fusion cRNA all produce fluorescent protein. In cells expressing ␣ 1 -EGFP alone and ␣ 1 -EGFP␥ 2, the fluorescent fusion protein appears to be distributed throughout the cell; whereas in cells expressing ␣ 1 -EGFP␤ 2 and ␣ 1 -EGFP␤ 2 ␥ 2, the protein seems to localize to the surface plasmalemma. These results suggest that the presence of a GABA A receptor ␤ subunit is needed for efficient translocation of ␣ 1 -EGFP to the cell surface.
fusion protein, when expressed in combination with the ␤ 2 or ␤ 2 ␥ 2 subunits, is preferentially localized to the surface plasmalemma with EGFP facing extracellularly. Since EGFP is fused to the carboxyl terminus of the ␣ 1 subunit, these results directly demonstrate that the COOH terminus of the ␣ 1 subunit is extracellular.
GFP-tagged GABA A Receptors Are Functionally Unaltered-To test whether the ␣ 1 -EGFP fusion protein could assemble into a functional GABA A receptor, oocytes expressing the fusion protein were tested with a two-electrode voltageclamp for the ability of GABA to activate a Cl Ϫ current. The traces in Fig. 4 (top) show Cl Ϫ currents activated by increasing concentrations of GABA from oocytes expressing ␣ 1 -EGFP␤ 2 and ␣ 1 ␤ 2 GABA A receptors. Fig. 4 (bottom) plots the GABAactivated current for ␣ 1 ␤ 2, ␣ 1 -EGFP␤ 2, ␣ 1 ␤ 2 ␥ 2 , and ␣ 1 -EGFP␤ 2 ␥ 2 receptors as a function of GABA concentration. The GABA EC 50 values for ␣ 1 -EGFP-containing receptors are similar to wild-type receptors (Table I). Oocytes expressing ␣ 1 -EGFP␤ 2 ␥ 2 receptors were also tested for the ability of diazepam to potentiate the GABA-mediated Cl Ϫ current. As seen in Fig. 5, the EC 50 for diazepam potentiation of the GABA response for ␣ 1 -EGFP␤ 2 ␥ 2 receptors is similar to wild-type receptors (Table I). The maximum potentiation obtained at 1 M diazepam was somewhat lower for ␣ 1 -EGFP␤ 2 ␥ 2 receptors, but still within reported ranges (19,21). Hill coefficients were not significantly different between ␣ 1 -containing versus ␣ 1 -EGFPcontaining receptors (data not shown). Taken together, these results indicate that COOH-terminal labeling of the ␣ 1 subunit of the GABA A receptor with EGFP yields fully functional ion channels in Xenopus oocytes when expressed with ␤ 2 or ␤ 2 ␥ 2 subunits.
In oocytes expressing ␣ 1 -EGFP␤ 2 ␥ 2 receptors, the emission output of ␣ 1 -EGFP was measured before and during a bath application of GABA to determine whether agonist-dependent rearrangements in the GABA A receptor would induce changes in the fluorescence of ␣ 1 -EGFP. No change in fluorescence was measured during a 100 M GABA application (n ϭ 3). DISCUSSION Because the response properties of neurons are ultimately controlled by the types of ion channels and neurotransmitter receptors present at a particular synapse, a fundamental question in neurobiology is to understand the regulation and the processes involved in the assembly, expression, and subcellular targeting of these proteins. Since GABA A receptors are heteropentameric proteins assembled from five distinct subunit classes with multiple subtypes, this question is particularly complex (22). To address this, we created a fusion protein of the GABA A receptor ␣ 1 subunit with a naturally fluorescent protein, EGFP. We then characterized ␣ 1 -EGFP expression in Xenopus oocytes, alone or with other receptor subunit combinations, using microscopic, immunological, and electrophysiological techniques. Attachment of GFP to a protein allows rapid, noninvasive detection of the tagged protein in living cells and provides a powerful tool for studying protein trafficking. Ideally, the labeled protein should be relatively unperturbed by its fluorescent tag and exhibit little or no change in its behavior.
Attachment of EGFP to the COOH terminus of the GABA A receptor ␣ 1 subunit produces a bright green fluorescent protein when expressed in Xenopus oocytes. When expressed alone or in combination with the ␥ 2 subunit, ␣ 1 -EGFP shows a gener- FIG. 3. The COOH terminus of the GABA A receptor ␣ 1 subunit is extracellular. Surface labeling of nonpermeabilized oocytes expressing ␣ 1 -EGFP␤ 2 ␥ 2 receptors by anti-EGFP antibodies and Texas Red supports the four-transmembrane subunit topology model. Red and green fluorescence were monitored in separate channels (left) and simultaneously (right). The fluorescence co-localizes and suggests that the majority of ␣ 1 -EGFP expressed is associated with surface receptors. alized fluorescence, which is distributed throughout the cell (Fig. 1). Immunolabeling with an anti-GFP antibody and Texas Red established that when expressed alone or with ␥ 2 , the location of the ␣ 1 -EGFP fluorescence is intracellular. Thus, neither ␣ 1 -EGFP nor ␣ 1 -EGFP␥ 2 is transported to the cell surface. Heterologous expression of single and multiple combinations of different GABA A receptor subunits and the subsequent formation of functional surface receptors has proven controversial. Homomeric expression of ␣ 1 and ␤ 2 subunits produces functional surface receptor expression in some reports (23) but not in others (24,25). Similarly, expression of binary combinations of subunits (e.g. ␣␥ and ␤␥) have produced functional receptors in some cases (26,27), but not in others (24,25). Our results provide evidence that ␣ 1 subunits are not localized to nor inserted into the surface membrane in oocytes expressing ␣ 1 subunits alone or ␣ 1 ϩ ␥ 2 subunits. The inability of ␣ 1 and ␣ 1 ␥ 2 subunit combinations to reach the cell surface may reflect the inability of these subunits to oligomerize into pentameric receptors.
When expressed with ␤ 2 or ␤ 2 ␥ 2 subunits, ␣ 1 -EGFP is targeted to the surface membrane (Fig. 1). Electrophysiological analyses show that the ␣ 1 -EGFP fusion protein forms func-tional GABA A receptor channels when coexpressed with ␤ 2 or ␤ 2 ␥ 2 subunits, with EC 50 values for both GABA and diazepam similar to wild-type receptors (Figs. 4 and 5), confirming a surface membrane localization. In addition, these results indicate that the GABA A receptor ␤ 2 subunit is required for translocation of ␣ 1 -EGFP to the surface plasmalemma, and most likely plays an important role in both functional pentameric receptor assembly as well as membrane targeting.
GABA A receptor targeting studies, carried out in polarized Madin-Darby canine kidney cells, have mixed results. One study (28) demonstrated that ␣ 1 subunits expressed alone were targeted to the basolateral membrane and ␤ 1 subunits expressed alone to the apical membrane. Coexpression of ␣ 1 and ␤ 1 targeted both subunits to the apical membrane. Thus, even though both ␣ 1 and ␤ 1 subunits are capable of being subcellularly targeted, the ␤ 1 subunit when present regulates the targeting of the ␣ 1 subunit. In contrast, Connolly et al. (29) showed that ␣ 1 ␤ 1 coexpression in Madin-Darby canine kidney cells resulted in a nonpolarized surface distribution, and neither ␣ 1 FIG. 4. ␣ 1 -EGFP containing GABA A receptors are functional and display wild-type sensitivity to GABA. Two electrode voltage-clamp of oocytes expressing ␣ 1 -EGFP␤ 2 , ␣ 1 ␤ 2 , ␣ 1 -EGFP␤ 2 ␥ 2 , and ␣ 1 ␤ 2 ␥ 2 receptors. Oocytes were treated with a range of GABA concentrations, and dose-response curves were fit to the data. Parameters from the curve fits are shown in Table I. ␣ 1 ␤ 2 and ␣ 1 ␤ 2 ␥ 2 receptors (open symbols, solid lines); ␣ 1 -EGFP␤ 2 and ␣ 1 -EGFP␤ 2 ␥ 2 receptors (filled symbols, dashed lines).

TABLE I Summary of voltage-clamp results
Dose-response data for wild-type and ␣ 1 -EGFP-containing subunit combinations for GABA and diazepam potentiation of GABA-mediated Cl Ϫ current in Xenopus oocytes are tabulated. Two-electrode voltageclamp and data analysis was performed as described (see "Materials and Methods"). Shown are means Ϯ S.D. for EC 50 values calculated from dose-response data (Figs. 4 and 5) using Prism software. I max is the range of maximal GABA-gated currents measured, maximum potentiation is the fold increase over GABA-gated current at 1 M diazepam, and n represents the number of oocytes tested. NS, no significant potentiation measured; ND, not determined.  5. Diazepam potentiates ␣ 1 -EGFP containing GABA A receptors. Oocytes expressing ␣ 1 -EGFP␤ 2 ␥ 2 and ␣ 1 ␤ 2 ␥ 2 receptors were treated with a range of diazepam concentrations in the presence of 1 M GABA. Potentiation response ratios were calculated by dividing peak GABA current in the presence of diazepam by peak response to GABA alone and normalized. Parameters from the curve fits are displayed in Table I. The EC 50 for diazepam potentiation of the GABA response for ␣ 1 -EGFP␤ 2 ␥ 2 receptors is similar to wild-type receptors. ␣ 1 ␤ 2 ␥ 2 receptors (open symbols, solid lines); ␣ 1 -EGFP␤ 2 ␥ 2 receptors (filled symbols, dashed lines).
nor ␤ 1 subunits expressed alone were inserted into any surface membrane. In the same study, ␣ 1 ␤ 2 coexpression resulted in a basolateral surface distribution, and ␣ 1 ␤ 3 in an apical distribution which switched to a basolateral distribution over time. Thus, while it appears that the type of ␤ subunit may guide membrane targeting, it is less clear if it is a required element. In our study, we show that ␣ 1 subunits expressed alone are not targeted to nor inserted into the oocyte surface membrane and that the ␤ 2 subunit is required for targeting and assembling the ␣ 1 subunit into a functional cell surface receptor.
The transmembrane topology of a GABA A receptor subunit is predicted by hydrophobicity analysis to have an extracellular NH 2 -terminal domain, four putative membrane-spanning domains, and an extracellular COOH terminus (30). Experimental support of this model is not well established for the GABA A receptor family, and as demonstrated for the ionotropic glutamate receptor family (15,31), hydrophobicity analysis is not always correct. For GABA A receptor subunits, the extracellular placement of the amino-terminal end of the subunit has been verified by antibody labeling of intact cells (25) and mutagenesis of glycosylation sites (32). In vitro and in vivo phosphorylation studies position part of the peptide loop between M3 and M4 intracellularly (33,34). In this study, since EGFP is fused to the carboxyl terminus of the ␣ 1 subunit, GFP immunocytochemical labeling of nonpermeabilized oocytes expressing ␣ 1 -EGFP␤ 2 or ␣ 1 -EGFP␤ 2 ␥ 2 receptors provides direct evidence that the COOH terminus of the ␣ 1 subunit is extracellular (Fig.  3). Our results lend further support for the four-transmembrane subunit topology model.
The stoichoimetry of recombinant GABA A receptors is most likely 2␣,2␤,1␥ or 3␣,2␤ subunits (35)(36)(37). Thus, ␣ 1 -EGFPcontaining receptors presumably have two to three EGFP tags per receptor. Surprisingly, the EGFP tag is functionally silent even though it is much larger than most protein tags and, at 238 amino acid residues, is approximately half the size of a GABA A receptor subunit. Despite this, the addition of this relatively large protein to the COOH terminus of the ␣ 1 subunit does not appear to interfere or alter subunit folding, oligomeric assembly, membrane targeting, or the physiological properties of the receptor. Since attachment of EGFP has no obvious effect on the receptor, we speculate that the COOH-terminal regions and the adjacent M4 transmembrane regions of the ␣ 1 subunits are relatively far apart from each other in a receptor complex and are presumably not involved in agonist-induced conformational changes in the receptor. The lack of change in EGFP fluorescent emission measured during GABA activation of ␣ 1 -EGFP␤ 2 ␥ 2 receptors is consistent with this hypothesis.
Taken together, our data unequivocally demonstrate that COOH-terminal tagging of the ␣ 1 subunit of the GABA A receptor with EGFP yields fully functional ion channels in Xenopus oocytes when expressed with ␤ 2 or ␤ 2 ␥ 2 subunits. Our results provide evidence for the importance of the ␤ 2 subunit in efficient functional receptor assembly, expression and targeting to the surface plasmalemma. In addition, we provide the first direct experimental evidence that the COOH terminus of the ␣ 1 subunit is extracellular. This is the first report of successfully tagging GFP to a member of the ligand-gated ion channel superfamily, which includes neuronal and muscle nicotinic acetylcholine receptors, GABA A receptors, glycine receptors, and serotonin receptors and establishes the general usefulness of this approach for studying their assembly, transport, targeting, and clustering. It should now be possible to identify specific regions of the receptor essential for these processes by using GFP-tagged subunit chimeras.