High-level expression and purification of Cys-loop ligand-gated ion channels in a tetracycline-inducible stable mammalian cell line: GABA_A and serotonin receptors

Abstract

The human neuronal Cys-loop ligand-gated ion channel superfamily of ion channels are important determinants of human behavior and the target of many drugs. It is essential for their structural characterization to achieve high-level expression in a functional state. The aim of this work was to establish stable mammalian cell lines that enable high-level heterologous production of pure receptors in a state that supports agonist-induced allosteric conformational changes. In a tetracycline-inducible stable human embryonic kidney cells (HEK293S) cell line, GABA_A receptors containing α1 and β3 subunits could be expressed with specific activities of 29–34 pmol/mg corresponding to 140–170 pmol/plate, the highest expression level reported so far. Comparable figures for serotonin (5-HT_3A) receptors were 49–63 pmol/mg and 245–315 pmol/plate. The expression of 10 nmol of either receptor in suspension in a bioreactor required 0.3–3.0 L. Both receptor constructs had a FLAG epitope inserted at the N-terminus and could be purified in one step after solubilization using ANTI-FLAG affinity chromatography with yields of 30–40%. Purified receptors were functional. Binding of the agonist [³H]muscimol to the purified GABA_AR was enhanced allosterically by the general anesthetic etomidate, and purified 5-hydroxytryptamine-3A receptor supported serotonin-stimulated cation flux when reconstituted into lipid vesicles.

Introduction

The human Cys-loop neuronal ligand-gated receptor ion channel superfamily, each of which consists of five homologous subunits arranged centrosymmetrically around an ion channel, include the gamma-aminobutyric acid Type A receptors (GABA_AR), glycine receptors (GlyR), nicotinic acetylcholine receptors (nAChR), and serotonin Type 3 receptors (5-hydroxytryptamine-3A receptor, 5-HT_3AR).^1–4 These receptors are the target of many drugs including anticonvulsants, sedatives, tranquilizers, anti-emetics, Alzheimer's therapies, and general anesthetics.⁵ They have been extensively studied by electrophysiological techniques so that their pharmacology and kinetics are quite well understood. However, there is little direct structural information because they do not express well. Indeed, structures are generally derived from homology models based on a cryoelectron microscopy structure at 4 Å resolution of the abundantly available muscle type nAChR from Torpedo electroplax⁶ and from crystallographic structures of prokaryotic channels that share some but not all of the structural domains of neuronal receptors.^7–9

Because of concern about protein misfolding, glycosylation, membrane lipid composition, and aggregation when proteins are expressed in the higher yielding bacterial, yeast, or insect cells, eukaryotic expression in mammalian cells is often preferred.^10–13 Heterologous expression of Cys-loop channels in mammalian cell lines has been achieved but rarely in the quantities and purity required for structural studies. For example, the best success has come from the 5-HT_3AR, where efforts over decades^14,15 have lead to yields of 100 pmol/plate with a specific activity of 12–60 pmol/mg.¹⁶ However, in the GABA_AR, much lower specific activities (1–6 pmol/mg) have been reported.^17–19 In contrast, in the single subunit G protein-coupled receptor (GPCR) superfamily sufficient quantities of protein have been expressed in mammalian cells to allow structural information to be obtained in solution by electron paramagnetic resonance (EPR)²⁰ and nuclear magnetic resonance (NMR) spectroscopy²¹ and recently by crystallography.²² A promising approach has been the employment of an induction strategy analogous to that used in bacterial expression systems that allows the cells to grow normally to high density before the demand of high-level protein expression is imposed on them.²³ Using this approach, it was recently reported that β2-adrenergic receptors were expressed at levels of 220 pmol/mg of membrane protein.²⁴

The aim of this work was to establish highly productive heterologous expression of human neuronal Cys-loop receptors in mammalian cells and to develop a one step purification of them in a state that supports agonist-induced allosteric conformational changes. Specifically, we set out to test whether the tetracycline-inducible system, which has enabled high-level production of GPCRs,^23,24 could be used to enhance expression in a pentameric protein. We first applied this strategy to the homomeric 5-HT_3AR and then to the heteromeric GABA_AR containing α1 and β3 subunits. Using the human embryonic kidney cells (HEK293S) tetracycline-inducible system, the strategy was found to be successful in both cases, allowing the production of 10 nmol of receptors in about a liter of suspension in a bioreactor.

Results and Discussion

Construction and expression of stable, inducible cell lines

Initially, to test whether the tetracycline-induction strategy could be applied to a Cys-loop ligand-gated ion channel (LGIC) superfamily protein, we used an existing mouse construct¹⁵ kindly provided by Horst Vogel and Ruud Hovius (Institute of Biomolecular Sciences, Swiss Federal Institute of Technology, Lausanne) to create both a HEK293S cell line stably expressing 5-HT_3ARs and a HEK293S-TetR-inducible cell line. In suspension, the stable cells achieved a specific activity of 8 ± 4 pmol of [³H]GR65630 sites/mg of membrane protein compared with 49 ± 3 pmol/mg in the tetracycline-inducible system (specific activity always refers to membrane protein). This sixfold increase may be compared with a 12-fold enhancement in a similar test with the single subunit β2-adrenergic receptor.²⁴

Encouraged, we created a HEK293S-TetR-inducible stable cell line expressing FLAG-h5-HT_3AR-1D4. We evaluated 70 clones for tetracycline-induced expression by reverse transcription-PCR and western blot using anti-5-HT_3AR and monoclonal rho-1D4 antibodies. Of the dozen lines chosen for further evaluation, some grew too slowly to be useful and others did not produce receptors when membranes were assayed by ligand binding assay. We selected for evaluation in suspension culture three clones that expressed 245–315 pmol [³H]GR65630 sites/plate at specific activities of 49–63 pmol/mg. Optimal inoculation conditions and induction times were established in 250 mL spinner culture flasks (see Supporting Information). Following induction, receptor production increased for 17 h, reaching a peak at 24 h (∼2 nmol/L; 17 ± 9 pmol/mg) before falling by a third at 36 h. In the bioreactor, 3 L cultures were grown for 5–6 days and then induced for 24 h, giving improved yields of 9–34 nmol/L at a specific activity of 24–47 pmol/mg (Table I).

Table I.

Yields and Specific Activities of Expressed Receptors

	Adherent culture on 15 cm platesa			Suspension culture in 3.5 L bioreactor
	Specific activity (pmol/mg)	Yield (pmol/plate)	Plates to express 10 nmolb	Specific activity (pmol/mg)	Yield (nmol/L)	Liters to express 10 nmolb
FLAG-5-HT3AR-1D4	49–63	245–315	30–40	24–47	9–34	0.3–1
FLAG-GABAAα1β3R	29–34	140–170	60–70	14	3.2	3

Receptor concentrations are in pmol or nmol of [³H]GR65630 binding sites for 5-HT_3AR and pmol or nmol of [³H]muscimol binding sites for GABA_Aα1β3Rs corrected to saturation.

^aOne hundred and seventy-six square centimeters per plate.

^bFor comparison, 100 g of bovine cerebral cortex has 1.3 nmol of [³H]muscimol sites.²⁵

With the 5-HT_3AR, the main advantage of using tetracycline induction is not the membrane specific activity achieved but the yield per liter. Thus, Hovius et al.¹⁵ report that mammalian cells infected with Semliki Forest virus had similar specific activities (25–50 pmol/mg) but yields (3.6–7.2 nmol/L) were three to five times lower than in our tetracycline-inducible cells. In adherent culture, we achieved 245–315 pmol/plate in the tetracycline-inducible cells compared with a recent report of 70–114 pmol/plate in a stable HEK293 cell line.¹⁶

We next extended our test of the tetracycline-inducible system to a heteromeric protein, the GABA_A receptor containing the human α1 and β3 subunits. About 12 of 40 clones evaluated by reverse transcription-PCR and by western blot for equal expression of the two subunits were further evaluated for receptor yield using [³H]muscimol binding, which requires the presence of both the α and β subunit.²⁶ The best clones expressed GABA_AR at levels of 140–170 pmol [³H]muscimol binding sites/plate and specific activities of 29–34 pmol/mg (Table I). This compares with 7.8 ± 0.5 pmol/mg for GABA_Aα1β2 receptors in Sf9 cells infected with baculovirus.¹⁷ For the suspension culture in the bioreactor, the production peaked 36 h after induction at 3.2 nmol of muscimol sites/L with a specific activity of 14 pmol/mg. Reeves et al.²³ found that addition of 5 mM sodium butyrate at induction was beneficial. We found its omission lowered the specific activity threefold, without altering the number of cells per liter.

In contrast to the 5-HT_3AR, tetracycline-induction led to the highest specific activities reported to date in cell membranes for GABA_ARs. In adherent culture, at ∼30 pmol/mg (Table I), the specific activity is some 50-fold higher than in brain²⁵ and nearly an order of magnitude higher than those reported for GABA_A α1β2γ2 receptors expressed in Ltk cells.¹⁸ Previous attempts using suspension cultures of Sf9 cells infected with baculovirus gave a specific activity of 2–8 pmol/mg and of 8 pmol/mg for GABA_A α1β2¹⁷ and α1β3¹⁹ receptors, respectively, compared with a value of 14 pmol/mg for tetracycline-induction in suspension in our work (Table I). In terms of overall yields, mammalian cells in suspension gave 3.2 nmol/L, comparable with 2.4 nmol/L in the insect cells above.¹⁹ The additional subunit in gamma-aminobutyric acid type A receptor with α1 and β3 subunits (GABA_Aα1β3R) did not result in markedly lower yields when compared with the 5-HT_3AR (Table I). To obtain the same number of GABA_ARs as available from 100 g of bovine cerebral cortex now requires eight to nine plates or 0.4 L culture.

Electrophysiological and pharmacological characterization of expressed receptors

Before progressing further, we established that the modified receptors behaved normally by electrophysiological criteria. FLAG-h5-HT_3A-1D4 receptors expressed in the HEK293S-TetR-inducible stable cell line had similar 5-HT concentration-activation curves, and rates of activation, deactivation, and desensitization as h5-HT_3AR expressed in HEK293 cells (Supporting Information, Fig. 1S)

Similarly, in the HEK293S-TetR-inducible stable cell line expressing FLAG–GABA_Aα1β3Rs, the concentration dependence with which GABA_A activated chloride currents was comparable with wild type GABA_Aα1β3Rs, as was the ability of etomidate to enhance currents elicited by low concentrations of GABA and, at higher concentrations, to activate currents in the absence of GABA²⁷ (Supporting Information Fig. 2S). Significantly, these currents were robustly inhibited by zinc chloride, a property consistent with the lack of a γ-subunit.²⁸ Desensitization and deactivation occurred with similar kinetics to the wild type receptor.

We compared the binding of various serotonergic ligands to human wild type and FLAG-5-HT_3A-1D4 receptors. The purification tags did not markedly change the binding affinity of the antagonist [³H]GR65630 or the agonist, 5-HT (Table II). The antagonists, ondansetron, and quipazine had dissociation constants and Hill coefficients of: 20 ± 1.3 and 1.2 ± 0.09, and 14 ± 1.2 and 1.0 ± 0.08, respectively, in the range of those reported in the literature.^15,29 Similarly, the muscimol and GABA binding parameters (Table II) of FLAG–GABA_ARs did not differ from those reported in the literature.^30,31

Table II.

Ligand Binding Parameters in Membranes and Reconstituted Receptors

Receptors	State	Ligand	K (nM)a,b	Hill coefficient
Wild type–h5-HT3AR	Cell membranec	[3H]GR65630	0.52 ± 0.02a	1.1 ± 0.1
	Cell membranec	5-HT	4100 ± 330b	1.1 ± 0.1
FLAG-h5-HT3AR-1D4	Cell membranec	[3H]GR65630	0.68 ± 0.04a	1.2 ± 0.1
	Cell membranec	5-HT	2200 ± 130b	0.9 ± 0.1
	Reconstituted in lipid bilayersde	[3H]GR65630	1.8 ± 0.2a	1.8 ± 0.3
	Control in cell membranesd	[3H]GR65630	1.8 ± 0.1a	1.8 ± 0.2
FLAG-hGABAAα1β3R	Reconstituteddf	GABA	64.1 ± 32.8b	0.8 ± 0.3
	Control in cell membranesd	GABA	62.3 ± 17.4b	0.5 ± 0.1
	Control in cell membranesc	[3H]muscimol	9.3 ± 2.4a	1.3 ± 0.5

^aK_d (equilibrium dissociation constant of the ligand) determined by titrating [³H]ligand.

^bK_i for ligands determined by displacement of 0.15 nM [³H]GR65630 or 2 nM [³H]muscimol (13% and 11% receptor occupancy, respectively). Inhibition constant (K_i) was calculated according to the equation K_i = IC₅₀/(1 + (ligand concentration/K_d ligand)). The half maximal inhibitory concentration (IC₅₀) was determined by nonlinear least squares fitting to the Hill equation.

^cFilter assay.

^dPrecipitation assay used for reconstituted receptors.

^eBilayer composition PC:PA:Chol, molar ratio 67:12:21.

^fReconstituted in 11.5 mM cholate, 0.86 mM asolectin.

Solubilization

Maximizing the solubilization of receptors from membranes is important because it is the first step in purification. It has been established that the detergent C₁₂E₉ [critical micelle concentration (CMC) 0.05 mM; aggregation number ∼142^32,33] is optimal for solubilizing the 5-HT_3AR, although its use is limited at higher concentrations by the irreversible loss of binding sites.¹⁵ We hypothesized that this loss could be avoided if the number of lipids in each micelle were maintained above a threshold, because it is known that in Torpedo nAChRs the number of lipids per oligomer must be maintained above ∼45 to avoid irreversible loss of binding sites.^34,35 Hovius et al. reported that at a fixed concentration of 1 mg membrane protein/mL, C₁₂E₉ concentrations ≥4 mM caused increasing loss of 5-HT_3AR binding sites.¹⁵ Under these conditions, we estimated that there were ≤30 lipids per micelle assuming equal amounts of protein and lipid by weight in the cell membranes. Consistent with our hypothesis, we found that by increasing the membrane protein concentration, and hence the number of lipids per micelle, we could safely raise the C₁₂E₉ concentration above 4 mM. In general, we also found it advantageous for solubilization to have about 2–10 micelles per GR binding site and ≥150 lipids per micelle. For example, at 20 mM C₁₂E₉ and 35 mg of membrane protein/mL (∼200 lipids per micelle), no binding sites were lost. This may explain the empirical observation that adding lipid during solubilization increases yields.¹⁴ The above solutions were rather viscous, so for convenience our standard conditions were 1 mM C₁₂E₉ at 1–2 mg/mL (estimated ∼150 lipids per micelle), which solubilized 80% ± 20% of the [³H]GR65630 binding sites from the cell membrane fraction. This low starting detergent concentration was advantageous during concentration by ultrafiltration when the detergent concentration rises.

For solubilization of brain GABA_AR, CHAPS (CMC 8 mM³⁶) has been used extensively,^25,37,38 but we found that GABA_ARs grown in HEK293S cells were poorly solubilized by it (Fig. 1), nor was this improved by adding asolectin to mimic the large amount of lipid present in brain. Recently, good solubilization of brain GABA_AR was reported with a combination of CHAPS and C₁₂E₉ (4:5 molar ratios).²⁵ Neither C₁₂E₉ alone nor plus CHAPS were very efficacious (Fig. 1). However, 2.5 mM n-dodecyl-β-d-maltopyranoside (DDM) alone solubilized 86% ± 18% of [³H]muscimol sites overnight and 62% ± 11% in 2 h.

Figure 1.

n-Dodecyl-β-d-maltopyranoside (DDM) is the best detergent for solubilizing the FLAG-hGABA_Aα1β3R from HEK293S-TetR cell membranes. The number of [³H]muscimol sites from membrane protein (1 mg/mL) solubilized increased with detergent concentration to a plateau and decreased thereafter. The low-CMC detergents, DDM (filled squares), and C₁₂E₉ (filled circles) solubilized best. Others detergents used were: C₁₂E₉:CHAPS (4:5 molar ratio) (filled triangles); CHAPS (open circles), and CHAPS supplemented with 1.8 mM asolectin (open triangles).

Affinity purification

We inserted the FLAG epitope following the putative signal sequence so that after processing it might be at the N-terminus and thus be recognized by the anti-FLAG M1 antibody, which is Ca²⁺ dependent, allowing convenient elution with EDTA rather than the FLAG peptide. In contrast, anti-FLAG M2 antibody binds the FLAG epitope in any position. FLAG–GABA_ARs bound equally to beads with either antibody. However, on anti-FLAG M1 agarose beads, EDTA at 2 and 10 mM eluted only 8% and 12% of bound receptor. Subsequent elution with the FLAG peptide (0.1 mM) eluted a further 33% of the retained receptor. We note that others have apparently found it necessary to add FLAG peptide during elution from M1 columns.³⁹ Solubilized FLAG-h5-HT_3AR-1D4 bound only to the anti-FLAG M2 agarose and were not retained by the monoclonal antibody rho-1D4 coupled to CNBr-activated Sepharose. Thus, we used the anti-FLAG M2 agarose to purify both receptors. The beads had capacities per milliliter of 4.4 nmol of [³H]GR65630 binding sites for 5-HT_3ARs and 1 nmol of [³H]muscimol binding sites for GABA_AR. Beads were equilibrated overnight with saturating concentration of receptors, washed and eluted in batch mode by equilibrating for 10 min with 0.1 mM FLAG peptide in the lipid–detergent buffer. Typically, three to four elutions of two bed volumes gave maximum recovery. Overall, about 50–70% of the receptor applied on the column was specifically eluted (Table III).

Table III.

Typical Purifications of 5-HT_3A and GABA_A Receptors by FLAG Affinity Chromatography

Receptor	Purification step	Total receptor (nmol)a	Total protein (mg)	Specific activity (pmol/mg)b	Yield (%)c	Average yield (%)d
FLAG-h5-HT3A-1D4	Membranes	34	700	49	100	100
	Solubilized membranes	25	357	68	73	86 ± 16
	Eluted receptor	18	–	–	54	54 ± 13
	Reconstituted receptor	13	5.5	2,300	38	34 ± 7
FLAG-hGABAAα1β3	Membranes	11	324	34	100	100
	Solubilized membranes	8.1	172	47	74	69 ± 13
	Eluted receptor	3.9	1.4	670	35	37 ± 2

^aReceptor concentration estimated as in Table I.

^bFLAG peptide is present in the eluted and purified preparations so that the specific activity given is underestimated by approximately a factor of two (see text).

^cYields are expressed relatively to the starting cell membrane preparation.

^dFigures are mean ± SD for four experiments for 5-HT_3AR purification and three experiments for GABA_AR purification.

Detergent removal and functional reconstitution

Because C₁₂E₉ and DDM have low CMC (50 μM³³ and 170 μM⁴⁰, respectively), they are difficult to remove during reconstitution. Therefore, for reconstitution studies, the column was washed and equilibrated with 10–23 mM cholate (CMC ∼14 mM⁴¹) and lipids. The receptor concentration decreased in successive elutions, which were dialyzed separately to produce batches of reconstituted receptor of increasing lipid:protein ratio.

Tracer amounts of [³H]cholic acid were added to a dialysis bag containing a parallel sample lacking receptors to monitor detergent removal. The time constant for removal was 4 h, meaning that 95% of the detergent was removed in 12 h (Supporting Information Fig. 3S). Dialysis was routinely continued for 1.5 days. After dialysis, 5-HT_3AR reconstituted in liposomes was collected by ultracentrifugation. About 50% of the GR65630-binding activity initially loaded onto the affinity column was recovered in the final pellet making total purification efficiency about 40% (Table III).

Once the FLAG–GABA_Aα1β3 receptors were bound to the anti-FLAG M2 column, DDM was exchanged for 0.86 mM asolectin in 11.5 mM cholate, conditions similar to those successfully used for the reconstitution of the GABA receptor from cow brain.²⁵ About 50% of the muscimol-binding activity loaded onto the column was recovered in the three eluates. Total purification efficiency averaged 40% (Table III). This preparation was functional (see below) and further reconstitution was not investigated.

Characterization of purified receptors

Because the eluting FLAG peptide is very difficult to remove, it was impossible to measure the specific activity of the purified receptors. Our value for the 5-HT_3AR of 2.3 ± 0.78 nmol/mg for a representative purification in Table III, therefore, represents a lower boundary of the specific activity. To obtain a rough estimate of the actual specific activity, we did a parallel dialysis of the complete elution buffer without any protein. This reduced the FLAG peptide's concentration from 100 μM to 40 μM. An estimate of 4 ± 3 nmol/mg was obtained using this value to correct the above specific activity. This estimate is closer to the claimed stoichiometry of one GR65630 site per pentamer¹⁵ than to the possible stoichiometry of 5, which would have a specific activity of 18.1 pmol/mg. This calculation assumes the receptor is pure (the high molecular weight bands in Figure 2 are positive in western blot using anti-5-HT_3AR and rho-1D4 antibodies). For the GABA_AR, we obtained a specific activity of 0.67 ± 0.18 nmol [³H]muscimol binding sites/mg, corresponding to 2.7 ± 1.02 nmol/mg after correction for the FLAG peptide, compared with 1.1 nmol/mg for the same receptor expressed in Sf9 cells and purified by Ni-NTA agarose¹⁹ and 3.4 nmol/mg for receptors purified from bovine cerebral cortex.²⁵ The calculated specific activity for pure receptors is 7.4 pmol/mg for two muscimol binding sites per oligomer. The reasons for such small discrepancies have been thoroughly discussed recently.^19,24

Figure 2.

Purification and deglycosylation of FLAG-h5-HT_3A-1D4 (lanes A and B) and FLAG-hGABA_Aα1β3 (lanes D and E) receptors. FLAG-h5-HT_3AR-1D4 (25 pmol/lane): silver stained SDS-PAGE gel (8 %) of the purified, reconstituted receptor (lane A), and after digestion with N-glycosidase F (lane B). The broad band between the 55 and 72 kDa markers is the receptor. After deglycosylation, it becomes a narrow band at 55 kDa. The band at 35 kDa originated from the enzyme. The high M_w bands are aggregated receptors (see text). FLAG-hGABA_Aα1β3R (18 pmol/lane): Coomassie blue stained SDS-PAGE gel (8%) of the purified receptor (lane D) and after digestion with N-glycosidase F (lane E). The presence of both, α1 and β3 subunits (the two diffuse bands at ±55 kDa marker) was confirmed via western blot and mass spectrometry. Deglycosylation produced two sharper, well-separated subunit bands at lower M_w. The gel was run out to increase the separation of the subunits. In regular gels (not shown), there were no contaminants below 40 kDa. Lanes C and F: molecular weight standards.

The purity of the reconstituted receptors was tested by SDS-PAGE (Fig. 2). For the 5-HT_3AR, the main band was observed at 55–72 kDa with additional staining at 110 and 135–170 kDa. The presence of 5-HT_3AR in all three regions was established by mass spectrometry. The higher M_w area represents aggregated receptors. In low-lipid reconstitutions, aggregation was reduced. The 110-kDa band also contained the ER-associated membrane chaperone calnexin (M_w 67,568 Da). Thus, this band, which was not observed routinely, is likely a complex of calnexin and 5-HT_3AR. This indicates that receptors are being processed in the endoplasmic reticulum, because calnexin is a calcium-binding protein that interacts with newly synthesized glycoproteins in the endoplasmic reticulum.⁴² Glycosylation provides further evidence of processing. N-glycosidase F caused the wide band at 55–72 kDa to move to lower M_w (∼55 kDa) and become sharper. Complete N-glycosylation of murine 5-HT_3AR expressed in COS-7 cells was shown to be essential for receptor assembly, plasma membrane targeting, ligand binding, and Ca²⁺ influx.⁴³

The GABA_AR appeared as two overlapping bands at ∼55 kDa with a smear of density at higher molecular weights (Fig. 2). Both alpha and beta subunits were detected in these bands by mass spectrometry and western blot. Deglycosylation resulted in two sharp better-resolved bands at lower molecular weight (predicted subunit M_ws, including purification tags but not the leader sequences are for α1 and β3 49.8 and 51.5 kDa, respectively).

Reconstitution of functional properties of expressed purified receptors

To examine the ability of purified reconstituted receptors to undergo ligand-induced conformational changes, we tested the ability of the GABA_AR to undergo conformational changes using the efficacy with which etomidate enhances [³H]muscimol binding.²⁵ However, enhancement of serotonin binding by anesthetics even in brain receptors is too modest to allow a similar assay.²⁹ Therefore, we fully reconstituted the latter receptor into defined lipid bilayers so that the ability of agonists to open the channel could be assessed by ⁸⁶Rb⁺ flux assay. Because of concern about particle size and variable high-nonspecific binding during filtration, for purified samples we used a poly(ethylene glycol) (PEG)-precipitation method for ligand binding assays (see Materials and Methods section).

When the FLAG-5-HT_3AR-1D4 was purified and reconstituted into lipid bilayers of defined composition (PC:PA:Chol, molar ratio 67:12:21), the dissociation constant of [³H]GR65630 was unchanged relative to the same receptor in cell membranes using a precipitation assay (Table II). These receptors also supported 5-HT stimulated ⁸⁶Rb⁺ efflux. In these vesicles about 0.01% of the total volume contained trapped ⁸⁶Rb⁺, half of which was released by 5-HT (Fig. 3). The 5-HT-stimulated influx was inhibited by ondansetron (100 μM), a competitive antagonist, and by picrotoxinin (1 mM), which is a channel blocker of 5-HT_3ARs.⁴⁴ Controls showed that 5-HT did not cause receptor-free vesicles to leak nonspecifically. Furthermore, the fraction of released ⁸⁶Rb⁺ decreased as the lipid:receptor ratio was increased, suggesting that ions not released by 5-HT are trapped in vesicles that do not contain receptors.

Figure 3.

Serotonin-stimulated efflux from ⁸⁶Rb⁺ loaded reconstituted FLAG-h5-HT_3AR-1D4 lipid vesicles. Vesicles (PC:PA:Chol, molar ratio 67:12:21) were loaded overnight at 4°C with ⁸⁶Rb⁺ before being passed down a Sephadex column containing either buffer (filled circles) or 100 μM 5-HT (open circles) to separate external ⁸⁶Rb⁺ from that retained in vesicles, which eluted in the void volume at 2–4 min. Vesicles in the presence of 5-HT released about half their trapped ⁸⁶Rb⁺ (see Results section). The inset shows the whole chromatogram, which is dominated by the external ⁸⁶Rb⁺ eluting at ≥6 min in equal quantities for both samples.

Preservation of function proved less robust in purified, reconstituted FLAG–GABA_Aα1β3Rs. Enhancement of [³H]muscimol binding by 10 μM etomidate was preserved when purified, reconstituted receptors were stored frozen and then thawed, but was lost as a function of time at 4°C at a rate of from 50% to 100% at 24 h. In contrast, the solubilized receptors retained their functionality for more than 24 h. The problem of stability in heterologously expressed and reconstituted proteins is currently an active area of research.^24,45 Subsequent assays were performed at 4°C at fixed time (4 h) after thawing the previously prepared samples.

In membranes, the binding of the agonist [³H]muscimol to FLAG–GABA_Aα1β3Rs was enhanced in a concentration-dependent manner by etomidate, attaining a maximum value of 180% by 10 μM before declining (Fig. 4). This is comparable with the behavior reported in the literature.^25,46 When the membranes were solubilized in 2.5 mM DDM, the enhancement curve was unchanged. However, in 11.5 mM cholate and 0.86 mM asolectin, higher etomidate concentrations were required to cause enhancement, possibly because the higher lipid concentration depleted the free concentration of etomidate, which is very lipid soluble.⁴⁷ This factor should decline in importance as the concentration of etomidate increases, and, indeed, at ≥30 μM etomidate enhancement was similar in all samples. When compared at 10 μM etomidate, however, enhancement was somewhat lower in the reconstituted samples than in the membranes (P = 0.033, unpaired t-test).

Figure 4.

Purified FLAG-GABA_A-α1β3Rs undergo allosteric conformation changes. Etomidate modulates specific [³H]muscimol binding (2 nM) in membranes (open diamonds with standard deviation (SD) as lines), in solubilized membranes (2.5 mM DDM, open circles with SD not shown for clarity), and after elution from the affinity column (11.5 mM cholate, 0.86 mM asolectin, filled squares with SD shown as capped lines).

Conclusion

We have demonstrated that two members of the Cys-loop LGIC superfamily, the 5-HT_3AR and GABA_AR, can be produced in much higher yields than previously reported by using an induction strategy in HEK cells. This extends the application of this methodology from the GPCR superfamily, to a five-subunit ion channel superfamily. Furthermore, the strategy was successful even when one of the receptors contained two different subunits. In membranes, the expressed receptors functioned as expected for native receptors by both electrophysiological (intact cells) and ligand binding criteria (membrane preparation). Thus, assembly and processing are normal by these criteria. We were able to purify the receptors in one step by conventional FLAG-affinity chromatography, and to establish conditions that preserved function as judged by the ability of ligands to cause conformational changes. The quantity and purity of these receptors is sufficient to allow application of techniques such as photolabeling to locate drug binding sites and spectroscopic methods such as EPR, fluorescence, and possibly NMR to characterize structure.

Materials and Methods

Materials

Synthetic oligonucleotides were purchased from MGH–DNA core facility (Boston, MA). Restriction enzymes and buffers were purchased from New England Biolabs (Ipswich, MA) or Stratagene (La Jolla, CA). HEK293S-TetR cells were a gift from H. G. Khorana's Laboratory at the Massachusetts Institute of Technology. Anti-FLAG M1 and anti-FLAG M2 affinity gels, FLAG peptide, asolectin from soybean, γ-aminobutyric acid, and quipazine were purchased from Sigma–Aldrich (St. Louis, MO). CNBr-activated Sepharose was from GE Healthcare Bio-Sciences (Piscataway, NJ). Detergents C₁₂E₉, sodium cholate (anagrade), DDM (anagrade), and CHAPS (anagrade) were from Anatrace-Affymetrix (Santa Clara, CA). Polar lipids from Avanti Polar Lipids (Alabaster, AL) were as follows: 16:0–18:1 PC (1-palmytoil-2-oleoyl-sn-glycero-3-phosphocholine), 16:0–18:1 PA (1-palmytoil-2-oleoyl-sn-glycero-3-phosphate, sodium salt), and cholesterol. [³H] GR65630 (75.7 or 83.8 Ci/mmol), [³H]muscimol (3-hydroxy-5-aminomethylisoxazole, [methylene-³H(N)]) (25.5 or 36.6 Ci/mmol), and ⁸⁶Rb (>1 Ci/g) were from Perkin–Elmer (Waltham, MA). Cholic acid, [2,4-³H] (25 Ci/mmol) was from American Radiolabeled Chemicals (Saint Louis, MO). The monoclonal antibody, rho-1D4, was prepared by the Cell Culture Center (Minneapolis, MN) from a cell line provided by R. S. Molday (University of British Columbia, Vancouver, Canada). Monoclonal antibody anti-GABA_Aβ2,3R extracellular domain was from Millipore (Billerica, MA). Polyclonal antibody anti-GABA_Aα1R was from Imgenex (San Diego, CA). N-glycosidase F was from ProZyme (Hayward, CA) or New England Biolabs (Ipswich, MA). Protein concentration was measured by BCA protein assay kit from Thermo Fisher Scientific (Rockford, IL) with appropriate background correction for the presence of detergent and lipid.

Dextran sodium sulfate (M_r 6000–8000) was from MP Biomedicals (Solon, OH), primatone RL/UF was from Kerry Bio-Science (Norwich, NY). d-Glucose, glutaMAX, pluronic F-68, and penicillin-streptomycin were from Invitrogen (Carlsbad, CA). Fetal bovine serum [(FBS), premium] was from Atlanta biologicals (Lawrenceville, GA). All reagents were purchased from Sigma Chemicals Co. (St. Louis, MO) unless otherwise stated.

Buffers were as follows: Buffer A (lysis buffer): 10 mM HEPES, 1 mM EDTA (pH 7.4), protease inhibitors freshly added before use (10 μg/mL pepstatin, 2 μg/mL aprotinin, 10 μg/mL chymostatin, 10 μg/mL leupeptin, and 1 mM phenylmethylsulfonyl fluoride; Buffer B (purification buffer): 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 2 mM CaCl₂, 5 mM KCl, 5 mM MgCl₂, 4 mM EDTA, 20% glycerol, and protease inhibitors as in buffer A with appropriate detergent, see Results and Discussion; Buffer C (GR65630 binding assay buffer): 10 mM HEPES, 1 mM EDTA (pH 7.4); Buffer D (muscimol binding assay buffer): 10 mM phosphate buffer (pH 7.4), 200 mM KCl, and 1 mM EDTA; Flux-buffer: 1.95 mM NaH₂PO₄·H₂O, and 3 mM Na₂HPO₄·7H₂O (pH 7.0), 250 mM NaCl, 5 mM KCl, and 2 mM MgCl₂; phosphate-buffered saline buffer (PBS): 137 mM NaCl, 2.7 mM KCl, 1.8 mM KH₂PO₄, and 10 mM Na₂HPO₄ (pH 7.4).

Cloning and expression

Human brain total RNA (Clontech, Mountain View, CA) was used to clone h5-HT_3A receptor into the EcoRI site of the pcDNA4/TO expression vector including the following features: (1) FLAG tag (DYKDDDDK) on the N-terminus immediately after the predicted signal sequence (see Supporting Information); (2) a bovine rhodopsin octapeptide tag 1D4 (TETSQVAPA) at the C-terminus followed by the natural stop codon, and (3) a Kozak consensus sequence (GCCACCATGG) to the ATG start codon creating the final construct FLAG-h5-HT_3A-1D4.

The GABA_AR α1 and β3 subunit genes were inserted in separate plasmids with independent antibiotic selection genes. First, an expression vector for the α1 subunit, which was a gift from Paul Whiting, Merck, Neurosciences, Harlow, Essex, UK, was used to create a FLAG–GABA_Aα1 construct in the pcDNA4/TO vector, which has a Zeocin resistance gene. An N-terminal FLAG tag was placed after the predicted signal sequence of 27 amino acids. From human brain total RNA (Clontech, Mountain View, CA), β3 subunit, isoform 2, was cloned into hygromycin selectable pcDNA 3.1 modified by adding two tetracycline operators after the last nucleotide of the TATA box in the CMV promoter. The identity of each gene was confirmed by sequencing the entire coding region.

Creation of stable transfected HEK293S-TetR cell lines

Suspension-enabled HEK293S-TetR cells constitutively express the tetracycline repressor protein (TetR), which represses gene expression,²³ until tetracycline is added. Transfection, creation of tetracycline-inducible stable cell line, maintenance of suspension cell culture, induction, cell harvesting, and collection of the cell pellet is described in detail in Supporting Information.

Preparation of membranes

Cell pellets, stored at −80°C, were thawed and resuspended in buffer A, homogenized (Ultra Turrax T25, IKA Works, Wilmington, NC) on ice for 3 × 5 s at 30 s intervals and then with 20 strokes in a Kontes-Duvall (Kimble-Kontes, Vineland, NJ) glass tissue grinder. Cell homogenates were centrifuged at 45,000g for 30 min at 4°C and pellets resuspended in a small volume of buffer A. Homogenization, grinding, and centrifugation were repeated as above. The final pellet was homogenized by aspiration through a 21-gauge needle and expulsion through a 27-gauge needle thrice. Membrane preparations (∼5–10 mg/mL) were stored at −80°C.

Solubilization, purification, and reconstitution of receptors

Solubilization of membranes (1 mg/mL in buffer B) in an appropriate detergent (1 mM C₁₂E₉ for 5-HT_3AR; 2.5 mM DDM for GABA_AR) was achieved by stirring overnight at 4°C followed by ultracentrifugation at 100,000g for 30 min at 4°C.

To reduce nonspecific retention, anti-FLAG affinity purification resins were preincubated with 3 mg/mL of poly-d-lysine hydrobromide (rocker platform; ∼2 h at 4°C), washed with buffer B, mixed with the solubilized supernatant (above), and rocked for 2 h at 4°C before being transferred to a Bio-Rad Econo column.

To replace the low-CMC solubilization detergent with lipids and dialyzable high-CMC detergents, receptors bound to the column were first washed with five column volumes of buffer B plus low-CMC detergent and lipids and then equilibrated with rocking overnight at 4°C in the wash buffer. The detergents were as follows: for 5-HT_3ARs, 11.5 mM cholate, 0.38 mM lipid mixture containing PC:PA:Cholesterol in molar ratio 67:12:21; for GABA_ARs, 11.5 mM cholate, 0.86 mM asolectin. Following a final five column volume wash, the purified receptor was eluted by competition with 0.1 mM FLAG peptide in the same wash buffer as above.

Purified 5-HT_3ARs were reconstituted by removing detergent by dialysis against 4 L of buffer B (glycerol and detergent omitted) for 1.5 days with a daily buffer change. For effective detergent removal, Bio-Beads SM-adsorbent (Bio–Rad, Hercules, CA) were added to the dialysis buffer (5 g/4 L). Finally, the dialyzate was centrifuged (130,000g for 4 h at 4°C), the pellet resuspended in a little buffer B, and aliquots frozen in liquid nitrogen and stored at −80°C. The purified GABA_ARs were functionally reconstituted without dialysis.

Radioligand binding assays

5-HT_3AR and GABA_ARs (25 μg of membrane protein in 500 μL assay buffer C or D, respectively) were equilibrated with radio-ligands at 4°C for 1 h, followed by filtration on GF/B glass fiber filters (Whatman, Schleicher & Schuell, Maidstone, United Kingdom) that were pretreated in 0.5% w/v poly(ethyleneimine) for 1 h. After receptor application, filters were washed twice under vacuum with 7 mL of cold assay buffer, and dried under a lamp for 30 min. Next, they were equilibrated in Liquiscint (Atlanta, GA) and counted (Tri-Carb 1900, Liquid Scintillation analyzer, Perkin–Elmer/Packard, Waltham, MA). To establish a complete binding curve, the concentration of radio-ligand was varied over a wide range and the free concentration was corrected by subtracting the bound concentration from the total radio-ligand concentration. Alternatively, purified reconstituted receptors were assayed by a precipitation assay that is more suitable to analyze binding of lipophilic ligands in the presence of lipids.⁴⁸ Samples prepared as above were supplemented with PEG (15.5% w/v final concentration) 15 min before centrifugation at 9500g for 30 min. Tubes containing the air-dried pellet were incubated in Liquiscint cocktail overnight before counting.

The 5-HT_3AR was assayed in buffer C with the antagonist [³H]GR65630 and nonspecific binding was determined in the presence of 1 μM Quipazine. Solubilized or purified GABA_ARs were assayed in buffer D supplemented as described by Li et al.²⁵ The GABA_AR was assayed with the agonist [³H]muscimol and nonspecific binding was determined in the presence of 1 mM GABA. Because of variable high-nonspecific binding, sites were assayed at 12 nM [³H]muscimol, and corrected to saturation using the binding parameters reported in Table II.

Rb⁺ flux assays of reconstituted 5-HT_3AR

Serotonin-induced efflux of ⁸⁶Rb⁺ from 5-HT_3A receptors in lipid vesicles was measured at 4°C. Vesicles were resuspended in flux-buffer and equilibrated with ⁸⁶Rb⁺ for 12 h at 4°C, loaded on a Sephadex G-50 (Sigma–Aldrich, St. Louis, MO) column (30 cm × 0.5 cm) and eluted using the flux-buffer with or without 100 μM serotonin. Fractions of 120 μL were collected at 30 s intervals, the optical density measured at 280 nm, and assayed by scintillation counting.

Electrophysiology

Whole-cell voltage-clamp was used to record currents from induced HEK293-TetR cells expressing either FLAG–GABA_Aα1β3 receptors or FLAG-5-HT_3A-1D4 receptors. Cells were seeded on a glass cover slip and protein expression was induced with tetracycline (1 μg/mL) for 22–30 h before recordings. Further details may be found in Supporting Information.

Glossary

Abbreviations:

5-HT₃AR

5-hydroxytryptamine-3A receptor

CMC

critical micelle concentration

DDM

n-dodecyl-β-d-maltopyranoside

EPR

electron paramagnetic resonance

GABA_Aα1β3R

gamma-aminobutyric acid type A receptor with α1 and β3 subunits

GPCR

G protein-coupled receptor

GlyR

glycine receptor

HEK293

human embryonic kidney cells

LGIC

ligand-gated ion channel

nAChR

nicotinic acetylcholine receptor

NMR

nuclear magnetic resonance

PEG

poly(ethylene glycol)