Soluble, oligomeric, and ligand-binding extracellular domain of the human alpha7 acetylcholine receptor expressed in yeast: replacement of the hydrophobic cysteine loop by the hydrophilic loop of the ACh-binding protein enhances protein solubility.

The N-terminal extracellular domain (ECD; amino acids 1-208) of the neuronal nicotinic acetylcholine receptor (AChR) alpha7 subunit, the only human AChR subunit known to assemble as a homopentamer, was expressed as a glycosylated form in the yeast Pichia pastoris in order to obtain a native-like model of the extracellular part of an intact pentameric nicotinic AChR. This molecule, alpha7-ECD, although able to bind the specific ligand alpha-bungarotoxin, existed mainly in the form of microaggregates. Substitution of Cys-116 in the alpha7-ECD with serine led to a decrease in microaggregate size. A second mutant form, alpha7-ECD(C116S,Cys-loop), was generated in which, in addition to the C116S mutation, the hydrophobic Cys-loop (Cys(128)-Cys(142)) was replaced by the corresponding hydrophilic Cys-loop from the snail glial cell acetylcholine-binding protein. This second mutant protein was water-soluble, expressed at a moderate level (0.5 +/- 0.1 mg/liter), and had a size corresponding approximately to a pentamer as judged by gel filtration and electron microscopy studies. It also bound (125)I-alpha-bungarotoxin with relatively high affinity (K(d) = 57 nm), the binding being inhibited by unlabeled alpha-bungarotoxin, d-tubocurarine, or nicotine (K(i) = 0.8 x 10(-7) m, K(i) = 1 x 10(-5) m, and K(i) = 0.9 x 10(-2) m, respectively). All three constructs were expressed as glycosylated forms, but in vitro deglycosylation reduced the heterogeneity without affecting their ligand binding properties. These results show that alpha7-ECD(C116S,Cys-loop) was expressed in P. pastoris as an oligomer (probably a pentamer) with a near native conformation and that its deglycosylated form seems to be suitable starting material for structural studies on the ligand-binding domain of a neurotransmitter receptor.

These membrane glycoproteins, which mediate rapid chemical synaptic transmission, form either homo-or heteropentamers made up of homologous subunits that show significant similarities in amino acid sequence, transmembrane topology, and overall secondary, tertiary, and quaternary structure, implying a common evolutionary origin (2,3).
AChRs, which are classified into muscle and neuronal types, are the best characterized members of the LGIC superfamily. Muscle-type AChRs, found at the vertebrate neuromuscular junction and in fish electric organ, have a fixed stoichiometry ((␣1) 2 ␤1␥␦ in Torpedo and embryonic mammalian AChR or (␣1) 2 ␤1⑀␦ in adult mammalian AChR), mediate neuromuscular transmission, and are implicated in the autoimmune disease myasthenia gravis. Neuronal AChRs, widely distributed in the pre-, post-, and peri-synaptic nerve terminals of the central and peripheral nervous system (4,5), exist either as heteropentamers containing two or three ␣-subunits (␣2-6) plus 2 or 3 ␤-subunits (␤2-4) or as homopentamers (␣7-9), with ␣7 being the only human subunit known to form a homopentamer (5). Neuronal AChRs play key roles in various neuron-neuron interactions and are therefore involved in many functions (reviewed in Ref. 5) such as ganglionic transmission, modulation of the release of various neurotransmitters (␥-aminobutyric acid, acetylcholine, serotonin, glutamate, dopamine, and noradrenaline) (6), attention, learning, memory consolidation, arousal, sensory perception, the control of locomotor activity, pain perception, and body temperature regulation (7). They are, thus, implicated in a number of serious neurological disorders including Alzheimer's disease, Parkinson's disease, schizophrenia, depression, autism, and forms of epilepsy as well as nicotine addiction (8). Because of this, neuronal AChRs have attracted much recent attention.
The thorough understanding of AChR function and its manipulation for therapeutic approaches requires the elucidation of its structure at high resolution. Currently available structural information on AChRs has been mainly derived from elegant cryo-electron microscopy studies on two-dimensional crystals of Torpedo AChR, which can be purified in large amounts. Such studies on membrane-bound Torpedo AChRs have recently provided the 4.6-and 4-Å resolution structures, respectively, of the extracellular and transmembrane domains of the AChR (9 -11). However, no x-ray structure of any AChR or any other LGIC has yet been obtained, although very small, non-diffracting three-dimensional crystals of Torpedo AChR have been reported (12,13). Attempts by several investigators to obtain diffraction quality three-dimensional crystals of this large membrane protein have met with little success.
The use of hydrophilic AChR polypeptide fragments, rather than full-length subunits, seems to be a more realistic ap-proach to the expression of and structural studies on the AChR. The ECDs of AChR subunits, corresponding to the N-terminal ϳ210 amino acids of the subunits (14), are involved in ligand binding and are therefore of high importance. The ECD of ␣-subunits bears the major loops contributing to the binding sites for agonists and competitive antagonists (15), whereas loops from adjacent subunits also contribute to ligand binding (reviewed in Ref. 16). A functional ␣7⅐5-hydroxytryptamine 3 receptor chimera consisting of the ␣7 AChR ECD fused to 5-hydroxytryptamine 3 receptor transmembrane and cytoplasmic domains (17) displayed typical ␣7 AChR pharmacology, showing the following: (a) that the N-terminal ECD and a complementary C-terminal domain can fold autonomously; and (b) that the ␣7-ECD contains all of the structural elements contributing to the neurotransmitter binding site. High-resolution structural analysis has been performed on the molluscan glial cell AChBP (18), a structural and functional homologue of the ECD of LGIC subunits. The AChBP forms a stable homopentamer and has 24% sequence similarity with human ␣7-ECD and less similarity with ECDs from all the subunits of other nicotinic receptors (20 -24%) or other LGICs (15-18%). However, the high-resolution structure of a LGIC ECD is not yet available. The ECD of the ␣7 subunit, which is known to form homopentamers (19,20), seems to offer the most promising approach to the problem because it could provide a model of the whole (pentameric) ECD of an AChR. Several investigators have expressed ␣7-ECDs (21)(22)(23)(24), either alone or fused to another (soluble) protein; however, the recombinant proteins, although soluble, were produced either in the form of microaggregates or in minute amounts and were thus unsuitable for x-ray studies.
In this study we expressed the human ␣7-ECD as a glycosylated form in the yeast P. pastoris (25,26), which has previously been used successfully in our laboratory to produce a glycosylated, soluble, and ligand-and antibody-binding human ␣1 ECD (27). The recombinant ␣7-ECD was found to be mainly in the form of microaggregates. By replacing the normally unpaired Cys-116 with serine and exchanging the hydrophobic Cys 128 -Cys 142 loop with the corresponding hydrophilic Cysloop of the AChBP, we obtained a protein, ␣7-ECD(C116S,Cysloop), which was produced in quantities of ϳ0.5 mg per liter of culture, was water-soluble, and existed as an oligomer with a size corresponding approximately to that of a pentamer. Both the glycosylated and the in vitro deglycosylated forms bound the specific ligand 125 I-␣-Bgt equally well and with high affinity, whereas binding was inhibited by unlabeled ␣-Bgt, d-tubocurarine, and nicotine. These results suggest that the double mutated protein, which seems to have a near native conformation and can be produced in satisfactory amounts, is a suitable material for high-resolution structural studies.
pPICZ␣A-␣7-ECD(C116S)-The 59-base 5Ј-AATTCTTCTGGGCAT-TCCCAGTACCTGCCTCCAGGCATATTCAAGAGTTCCTGCTACAT-3Ј and the 57-base 5Ј-CGATGTAGCAGGAACTCTTGAATATGCCTGGA-GGCAGGTACTGGGAATGCCCAGAAG-3Ј oligonucleotides were synthesized to be complementary to each other. These oligonucleotides contain a single nucleotide change (double underlined) resulting in replacement of the Cys at position 116 of the ␣7-ECD with Ser. An equimolar mixture of the two oligonucleotides was annealed, and the resultant double-stranded DNA bearing the EcoRI and ClaI sites at its ends was substituted for the corresponding fragment in pPICZ␣A-␣7-ECD. (Fig. 1B, section b).

Protein Expression and Purification
Protein Expression in P. pastoris-In pilot experiments, single colonies of transformed cells were used to inoculated 2.5 ml of BMGY medium (1% (w/v) yeast extract, 2% (w/v) peptone, 100 mM potassium phosphate, pH 7.0, 1.34% (w/v) yeast nitrogen base, 4 ϫ 10 Ϫ5 % (w/v) biotin, and 1% (v/v) glycerol). After 16 -20 h of incubation at 30°C, the cells were resuspended in 5 ml of BMMY medium (BMGY medium in which the glycerol was replaced by 0.5% (v/v) methanol) to induce expression. Over the next 4 days, induction was maintained with the daily addition of 0.5% (v/v) methanol, and the culture supernatants were tested for expression of the recombinant proteins by dot-blot analysis using the anti-Myc mAb 9E.10 (American Type Culture Collection). The clone with the highest protein yield was selected for medium scale protein expression in 2-liter flasks. In all subsequent preparations the cells were harvested at ϳ24 h of methanol induction, as very little increase in the yield for any of the three products was seen after longer induction periods.
First Stage Protein Purification-The culture supernatant was concentrated 10-fold by the addition of solid NH 4 (SO 4 ) 2 to 60% saturation in the presence of 0.05% (w/v) NaN 3 , 1 mM phenylmethylsulfonyl fluoride, and 1 mM NaF followed by overnight precipitation at 4°C and dissolution of the pellet in the appropriate volume of distilled H 2 O. The concentrated protein solution was dialyzed against 50 mM PB, pH 8, and the recombinant protein was purified by metal affinity chromatography on Ni 2ϩ -NTA (Qiagen) according to the manufacturer's instructions. Elution was performed stepwise under native conditions with 50 mM PB, pH 8.0, containing 0.5 M NaCl and 50, 100, 150, or 1000 mM imidazole. The eluates were analyzed by 12% SDS-PAGE followed by Coomassie Brilliant Blue staining or Western blot analysis using either the anti-Myc mAb 9E.10 or rabbit polyclonal anti-␣7 antibodies. Anti-␣7 antibodies were produced by Dr. Leslie Jacobson in our laboratory by immunizing rabbits with either ␣7 synthetic peptide 176 -212 (anti-␣7 peptide polyclonal antibody) or Escherichia coli-expressed ␣7-ECD-(1-209) (anti-bacterial ␣7-ECD polyclonal antibody). The 100 mM imidazole eluate was used for in vitro deglycosylation, as the ␣7-ECD was eluted mainly at this fraction.
Gel Filtration Chromatographic Analysis and Second Stage Purification-After high speed centrifugation (1 min at 13,000 rpm) and filtration through a 2.2-m filter (Millipore), 1-ml samples of the Ni 2ϩpurified ␣7-ECDs were analyzed on an Ä KTA 90 purifier system (Amersham Biosciences) using a 24-ml Superose 12 column (Amersham Biosciences) and isocratic elution with 10 mM PB, pH 7.4, at a flow rate of 0.5 ml/min. 0.5 ml fractions were collected and analyzed for 125 I-␣-Bgt binding. The protein concentration of the gel-filtered material was determined spectrophotometrically based on the absorbance at 280 nm (A 280 ) and the protein content in aromatic residues (Trp and Tyr) and disulfide bonds (S-S) using the following formula: molarity ϭ (A 280 ϫ dilution)/(5550Trp ϩ 4910Tyr ϩ 125 S-S) (29).

125
I-␣-Bgt Binding Tests-Various amounts of Ni 2ϩ -purified or gel filter-purified protein were incubated with 50,000 cpm 125 I-␣-Bgt at 4°C, either for 1.5 h (␣7-ECD(C116S,Cys-loop)) or overnight (␣7-ECD or ␣7-ECD(C116S)) in a final volume of 50 l of PB-BSA (10 mM phosphate buffer, pH 7.4, containing 0.2% BSA). The samples were then diluted in 1 ml of Tris-Triton X-100 buffer and filtered immediately through two Whatman DE81 filters presoaked with Tris-Triton X-100 buffer; they were then washed twice with 1 ml of Tris-Triton X-100 buffer, and the bound radioactivity was measured on a ␥-counter. Samples without recombinant ECDs were used to measure nonspecific binding.
Competition Experiments and Scatchard Plots-To measure the specific binding of ␣-Bgt and to study the binding of other small nicotinic ligands such as nicotine, carbamylcholine, d-tubocurarine, epibatidine, acetylcholine, methyllycaconitine, or gallamine, only gel-filtered samples were used. For Scatchard plot analysis, various concentrations (40 -1500 fmol) of 125 I-␣-Bgt (specific activity 800,000 cpm/fmol) were incubated with 50 ng of recombinant protein at 4°C for 1.5 h or overnight as described above in a final volume of 50 l of PB-BSA, and the bound 125 I-␣-Bgt was measured using Whatman DE81 filters as described above. In competition experiments, various amounts of the unlabeled ligands were preincubated for 3 h at 4°C with 50 ng of the recombinant protein diluted in PB-BSA (final volume 40 l), then 10 l of 50,000 cpm 125 I-␣-Bgt in PB-BSA was added, and the mixture was incubated for 1.5 h or overnight as described above. Then the bound 125 I-␣-Bgt was measured using Whatman DE81 filters as described above. To measure inhibition of the initial rate of toxin binding, the competition experiments were run as described above except that the mixture was incubated in the presence of 125 I-␣-Bgt for only 10 min. Residual 125 I-␣-Bgt binding ability was determined as the percentage of the radioactivity bound in the presence of the unlabeled ligand compared with its absence. When measuring the inhibition by various drugs on the binding of 125 I-␣-Bgt to ␣7-ECD(C116S,Cys-loop), the IC 50 values were determined, and the K i was calculated using the formula, In Vitro Deglycosylation-Approximately 100 g of Ni 2ϩ -NTA purified ␣7-ECDs was deglycosylated by the addition of 500 units of Nglycosidase F (peptide:N-glycanase F; New England Biolabs) under native conditions without heating, denaturation, or the addition of any detergent. The reaction was performed in 50 mM PB, pH 7.5 (final volume 400 l), at 4°C (a much lower temperature than that suggested by the manufacturer) and incubated overnight (although the reaction was completed within 1-2 h). The resulting deglycosylated protein was detected on Western blots using anti-Myc mAb 9E. 10.
Electron Microscopic Analysis-A colloidin film bearing copper grids (400 mesh; TAAB Laboratories Equipment Ltd.) was coated with a thin carbon layer and glow-discharged for 2 min prior to use. Recombinant protein gel-filtered samples (1 g/l) were added to the carbon-coated grids and embedded in 2% (w/v) uranyl acetate as negative stain. Images were taken at 40,000ϫ magnification using a Philips 208 electron microscope operating at 80 kV (30).

RESULTS
Expression and Study of the ␣7-ECD-Recombinant ␣7-ECD was expressed as a soluble protein in the P. pastoris culture supernatant. This protein was concentrated by ammonium sulfate precipitation and dialyzed against 50 mM PB, pH 8.0; then ␣7-ECD was purified by affinity chromatography on a Ni 2ϩ -NTA column using the His 6 tag fused to its C terminus (Fig.  1C). Approximately 60% of the recombinant protein was eluted in the 100 mM imidazole fraction, and another 30% was eluted in the 150 mM imidazole fraction as judged by SDS-PAGE followed by staining with Coomassie Brilliant Blue ( Fig. 2A,  lanes 1 and 2) and Western blot using the anti-Myc mAb (Fig.  2B, lanes 1 and 2) and confirmed by 125 I-␣-Bgt binding experiments (data not shown). The eluted protein was not pure, with the main contaminants being two bands of ϳ40 and ϳ70 kDa (see Fig. 2A, lanes 1 and 2). The protein was further identified in Western blots by the anti-␣7 peptide polyclonal antibody and the anti-bacterial ␣7-ECD polyclonal antibody. The purified ␣7-ECD was able to bind 125 I-␣-Bgt in solution in filter assay experiments and was blocked by unlabeled ␣-Bgt or d-tubocurarine (data not shown).
The molecular mass of the product estimated by SDS-PAGE (Fig. 2, A and B, lanes 1 and 2) was ϳ36 -37 kDa, much higher than the value of 27.3 kDa predicted from its amino acid sequence. This difference was shown to be due to glycosylation, because in vitro deglycosylation of the Ni 2ϩ -purified protein with N-glycosidase (peptide:N-glycanase) reduced the apparent molecular mass on SDS-polyacrylamide gels to approximately that expected as shown by Western blot analysis (Fig. 3A, left pair of lanes). This shift indicates that, like the native ␣7 AChR, the recombinant protein was produced as a glycosylated form. The deglycosylated bands also appeared to be sharper, probably due to elimination of various glycosylated forms with different molecular masses. When the glycosylated and in vitro deglycosylated forms were tested for 125 I-␣-Bgt binding, no significant difference was seen in their binding efficiencies (Fig. 3B, left pair of bars).
The oligomeric state and size of the recombinant protein in detergent-free solution (10 mM PB, pH 7.5) were studied by gel filtration analysis on a Superose 12 column, the fractions being assayed for 125 I-␣-Bgt binding. The bulk of the protein was eluted close to the column void volume, indicating the presence of high molecular mass aggregates (Fig. 4A). In fact, precipitation of the protein usually occurred after 1-2 days of storage at 4°C in imidazole-containing 50 mM PB and 0.5 M NaCl.
Expression and Purification of Mutated Forms of ␣7-ECDs-In an effort to avoid microaggregate formation, we then constructed two mutants, i.e. the ␣7-ECD(C116S) in which Cys 116 was replaced by serine and the ␣7-ECD(C116S,Cysloop) in which, in addition to the C116S mutation, the hydrophobic Cys-loop was replaced by the corresponding hydrophilic Cys-loop from the snail AChBP. The mutated proteins were produced in P. pastoris and purified using Ni 2ϩ -affinity chromatography as described for ␣7-ECD. Again, the proteins eluted mainly in the 100 mM imidazole fraction, whereas a small percentage eluted in the 150 mM imidazole fraction ( Fig.  2A, lanes 3-6). The ϳ40 and ϳ70 kDa contaminant bands observed after ␣7-ECD Ni 2ϩ -NTA purification, making up ϳ50% of the protein, were also present in the metal-purified mutated protein preparations; these bands were not recognized either by the anti-Myc mAb (Fig. 2B) or by anti-␣7 polyclonal sera (data not shown) and were therefore attributed to P. pastoris intrinsic proteins. The mutated ␣7-ECDs were further purified by gel filtration on Superose 12, giving ϳ90 -95% pure protein as estimated by SDS-PAGE followed by Coomassie Brilliant Blue staining (Fig. 4C, inset). The purified mutated proteins were also recognized by the two anti-␣7 antisera described in the section on ␣7-ECD (data not shown). The yield of gel filtration purified ␣7-ECD(C116S) was ϳ0.2 mg per liter of culture compared with that of ϳ0.1 mg per liter for gel-filtered ␣7-ECD, whereas that of ␣7-ECD(C116S,Cys-loop) was ϳ0.5 Ϯ 0.1 mg per liter of culture. These yields were reproducible in several expression studies. ␣7-ECD(C116S), stored at 4°C in imidazole containing 50 mM PB and 0.5 M NaCl, precipitated after a short period of storage but to a somewhat lower extent than ␣7-ECD. In contrast, ␣7-ECD(C116S,Cys-loop) did not precipitate after storage for several months at 4°C with a concentration of ϳ0.3 mg/ml in the same buffer (50 mM PB, 0.5 M NaCl, and 100/150 mM imidazole). The molecular masses of the mutated proteins were estimated by SDS-PAGE to be the same as that of ␣7-ECD, i.e. ϳ36 -37 kDa, indicating that the mutants were produced as glycosylated forms.

Effect of in Vitro Deglycosylation of the Mutant Proteins on Molecular Mass and 125 I-␣-Bgt
Binding-Deglycosylation of the Ni 2ϩ -purified mutated proteins ␣7-ECD(C116S) and ␣7-ECD(C116S,Cys-loop) resulted in a sharper double band of approximately the expected molecular mass of 27.3 kDa (Fig.  3A). As with ␣7-ECD, the 125 I-␣-Bgt binding capacities of equal concentrations of the glycosylated and in vitro deglycosylated forms did not differ significantly (Fig. 3B).
Characterization of the Oligomeric State and Size of the Mutants by Gel Filtration Analysis-Gel filtration analysis on a Superose 12 column was used to study the effect of mutations on the oligomeric state and size of the recombinant proteins. The bulk of ␣7-ECD(C116S) eluted later than the void volume fractions but still at an apparent molecular mass much larger than those of pentamers (ϳ440 -300 kDa) (Fig. 4B). However, the double mutant ␣7-ECD(C116S,Cys-loop) eluted in fractions corresponding to molecular masses between 232 and 158 kDa (Fig. 4C), the peak fractions corresponding to an apparent molecular mass of ϳ210 kDa. This value is slightly higher than the expected molecular mass for the glycosylated ECD pentamer calculated using the apparent molecular mass of SDSpolyacrylamide gels (5 ϫ 36 -37 ϭ 180 -185 kDa). This small difference may be due to the shape of the oligomeric molecule.  (lanes 1, 3, and 5) or 150 mM imidazole (lanes 2, 4, and 6). The positions of the recombinant proteins (arrows) and of the molecular mass standards are indicated. All recombinant proteins had an apparent molecular mass of ϳ36 -37 kDa. Binding of 125 I-␣-Bgt and Cholinergic Ligands to Purified ␣7-ECD(C116S,Cys-loop)-The ability of the three gel filtration-purified ␣7-ECDs to bind 125 I-␣-Bgt in solution was tested by filter assay experiments as an indication of correct protein folding (Fig. 5). Although all were able to bind this specific ligand, the ␣7-ECD(C116S,Cys-loop) showed a much higher binding capacity. Moreover, 125 I-␣-Bgt binding to ␣7-ECD(C116S,Cys-loop) reached saturation at 45 min of incubation at 4°C, whereas for the other two constructs overnight incubation with 125 I-␣-Bgt was required to achieve saturation (not shown). The K d , for the gel-filtered double mutant, estimated by Scatchard analysis, was 57 nM (Fig. 6A).
The specificity of 125 I-␣-Bgt binding to gel-filtered ␣7-ECDs was demonstrated by competition experiments with various concentrations (1 nM-10 M) of unlabeled ␣-Bgt (Fig. 6B). In the case of the ␣7-ECD(C116S,Cys-loop) for which K d was estimated, the results of competition experiments (K i ϭ 1.5 ϫ 10 Ϫ7 M) were consistent with the K d estimated from Scatchard plot (K d ϭ 57 nm).
To further test the native-like conformation of the ␣7-ECD(C116S,Cys-loop), several known cholinergic ligands were tested in competition experiments for inhibition of 125 I-␣-Bgt binding to the gel-filtered protein. As shown in Fig. 6B, the competitive antagonist d-tubocurarine had an IC 50 of 3 ϫ 10 Ϫ5 M (K i ϭ 3 ϫ 10 Ϫ5 M). Interestingly, the agonist nicotine also inhibited 125 I-␣-Bgt binding, although with a much lower affinity (K i ϭ 1 ϫ 10 Ϫ2 M). The muscle AChR-specific antagonist gallamine had no effect on 125 I-␣-Bgt binding. The lack of effect of this ligand, even at high concentrations, shows that the inhibition caused by the other ligands was specific and that the ligand-binding site retains, at least to some extent, its ␣7 subtype specificity. However, no inhibition of 125 I-␣-Bgt binding was seen using the AChR ligands carbamylcholine, epibatidine, acetylcholine, or methyllycaconitine (data not shown). When the same inhibition experiments were performed at the initial rate of toxin binding, i.e. after 10 min incubation with 125 I-␣-Bgt, the K i values moderately improved. They were estimated to be 0.8 ϫ 10 Ϫ7 M, 1 ϫ 10 Ϫ5 M, and 0.9 ϫ 10 Ϫ2 M for ␣-Bgt, d-tubocurarine, and nicotine, respectively (Fig. 6C). Again, no inhibition was observed when carbamylcholine, epibatidine, or methyllycaconitine were used (not shown).
Electron Microscopic Evidence for Oligomeric Assembly of ECD Complexes-Electron microscopy of samples in negative stain was used to monitor the presence of complexes formed by the ECDs (kindly performed by Dr. N. Unwin, Cambridge, UK). The glycosylated and deglycosylated ␣7-ECD(C116S) was rather in the form of aggregates much larger than pentamers. The deglycosylated ␣7-ECD(C116S,Cys-loop) sample was quite homogeneous, consisting of oligomeric complexes resembling in size and shape the negative stained images of the native AChR (31) and the assembly of four muscle AChR subunit ECDs heterologously expressed in a baculovirus system (30). The characteristic donut shape could be seen, although not very clearly. The complexes showed a preferential "end-on" orientation, although side views of the ECD pentamers could also be seen (Fig. 7). DISCUSSION Nicotinic AChRs are the best studied members in the LGIC superfamily. In depth knowledge of their structure would facilitate the understanding of their function and the design of subtype-specific drugs. Despite numerous biochemical, biophysical, and cryo-electron microscopy studies, no high-resolution structure of these proteins is available. Even when sufficient protein amounts can be isolated, as in the case of the Torpedo AChR, crystallization of such large membrane proteins seems to be a difficult task mainly because of the large transmembrane (hydrophobic) region. The production of their N-terminal domains in heterologous expression systems appears to be a more realistic approach toward crystallization and high-resolution crystallographic studies, because these domains are extracellular and are therefore expected to be hydrophilic and, thus, watersoluble. Recombinant ␣7-ECD production would be appropriate for obtaining an oligomeric AChR ECD form.
Several attempts have been made to express ␣7-ECDs from different species in both prokaryotic and eukaryotic expression systems. Expression of chick ␣7-ECD in Xenopus oocytes resulted in a water-soluble (at least at the low concentrations tested) ligand-binding protein of a size probably corresponding to pentamers, but the protein was produced in only minute amounts (21). When sufficient protein is produced, as in the case of rat ␣7-ECD expressed in bacteria as a fusion protein with maltose-binding protein (22) or glutathione S-transferase (24), the main problem seems to be the formation of high molecular mass aggregates rather than monomers or small oligomers, probably due to incomplete folding of the protein and exposed hydrophobic regions.
In the present study, we used the P. pastoris eukaryotic system to express human ␣7-ECD. This system has been successfully used in our laboratory to express the human muscle AChR ␣1-ECD in a soluble, ligand-binding monomeric form (27). However, expression of human ␣7-ECD in this same yeast system resulted in the recombinant protein being produced mainly in the form of microaggregates, which, however, could bind 125 I-␣-Bgt and d-tubocurarine. To try to produce a more suitable material, we constructed two mutant forms of the protein.
The first mutation, designated ␣7-ECD(C116S), involved the replacement of Cys 116 by Ser. In native AChRs, Cys 116 is an unpaired cysteine that has been shown to be non-critical for the assembly and function (ligand binding ability) of the ␣7 subunit in the intact AChR (32). However, it may be involved in the formation of intermolecular bonds during ␣7-ECD expression, and its replacement by serine has been successfully used to produce a rat glutathione S-transferase-␣7 (1-208) ECD in a form which is more soluble than the unmodified protein (increased stability and decreased aggregation) (23). Moreover, as described above, the expression of ␣1 ECD, which does not contain this extra Cys, did not result in microaggregate formation (27). ␣7-ECD(C116S) bound 125 I-␣-Bgt somewhat better than did unmodified ␣7-ECD (Fig. 5), and gel filtration chromatography showed a shift of the elution peak to fractions corresponding to lower molecular masses, which were, however, still mainly larger than pentamers. However, a small fragment eluted in fractions corresponding approximately to the size expected for the pentamer.
Construction of the second mutation involved the amino acids of the Cys-loop (Cys 128 -Cys 142 ) in addition to the C116S point mutation. In pentameric LGICs, this Cys-loop is well conserved and mainly hydrophobic (CYIDVRWFPFDVQHC in human ␣7), whereas in AChBP (the only LGIC-related protein yet crystallized) it is more hydrophilic (CDVSGVDTESGATC). The crystal structure of the AChBP suggests that the Cys loop is located at the bottom of the ECD, close to the C terminus of this domain. This implies that the corresponding hydrophobic region in the LGICs could interact with the membrane or the transmembrane region of these receptors. Thus, it may be mainly buried in intact proteins but exposed in free ECDs and be a source of microaggregate formation. In fact, the Torpedo  californica ␣-subunit ECD fragment ␣143-210, which does not contain the Cys-loop, was expressed as a soluble form (33), further supporting the role of the Cys-loop in ECD solubility. Hoping to obtain a more soluble protein, we therefore constructed a mutant in which, in addition to the C116S mutation, the ␣7 hydrophobic Cys loop (15 amino acids) was replaced by the corresponding hydrophilic AChBP loop (14 amino acids). These amino acid residues do not seem to contribute to the ␣-Bgt-binding site (34).
This double mutant was expressed at higher levels than were the previous two molecules and was able to bind ␣-Bgt with higher efficiency. Its K d for ␣-Bgt binding was estimated to be 57 nM (Fig. 6). This K d value corresponds to an affinity ϳ44 times higher than that of the rat ␣7-ECD expressed in E. coli as a soluble fusion protein (K d ϭ 2.5 M) (22) and only 28 times lower than that of the native ␣7 AChR (K d ϭ 1.9 nM) (35). The fact that the ␣7-ECD(C116S,Cys-loop) was able to bind the agonist nicotine, even if with low affinity, is a strong indication for the presence of adjacent subunits, possibly generating the correct form of an AChR, i.e. a pentamer. Gel filtration chromatography showed that its size corresponded approximately to that of a pentamer. Further confirmation was obtained from electron microscopy analysis of deglycosylated gel-filtered ␣7-ECD(C116S,Cys-loop) by Dr. N. Unwin (Cambridge, UK), which showed that the molecule had the size and shape expected for a pentamer complex and that, in several cases, structures that look like donuts seen from the side, as well as end-on images viewed down the central pseudo-5-fold axis, could be observed.
In vitro deglycosylation of any of the three proteins did not affect their ability to bind 125 I-␣-Bgt. This lack of a significant difference in 125 I-␣-Bgt binding between the glycosylated and deglycosylated forms suggests that glycosylation is not crucial for ␣7 ligand binding. This contrasts with our earlier observation that deglycosylation of ␣1 ECD expressed in the same eukaryotic system dramatically reduced ␣-Bgt binding (27). The solved structure for the AChBP provides an explanation for this difference. Assuming that the threedimensional structures of the two ECDs (␣1 and ␣7) are similar to that of the AChBP (36), the ␣1-subunit glycosylation site (Asn 141 ) would be in close proximity to the ligandbinding site, explaining the dramatic effect of deglycosylation on ␣-Bgt binding. However, in the case of ␣7-ECD the three putative glycosylation sites (Asn 22 , Asn 68 , and Asn 111 ) would be quite far away from where the ligand-binding site resides and would be less likely to be involved in ligand binding. This assumption is further reinforced by our previous observations on human ␣1 and ␣7-ECDs expressed in E. coli as a non-glycosylated form. E. coli ␣1-ECD was not able to bind 125 I-␣-Bgt (37) whereas E. coli expressed ␣7-ECD was able to bind 125 I-␣-Bgt, 2 although with low affinity (34).
The observation that the deglycosylated form binds ligands to the same extent as the glycosylated form may be very useful for crystallization studies, as it means that either form could be used for structural studies on the ␣7 ligand-binding domain. Deglycosylated ␣7-ECD(C116S,Cys-loop) seems more promising as a starting material for structural studies because glycosylation is a source of heterogeneity, which is a very undesirable feature in terms of crystallization. However, the use of a eukaryotic system for the initial production of the glycosylated form, followed by its in vitro deglycosylation, may be required to produce the molecule in a native-like form, because AChR subunit glycosylation has been reported to be crucial for the correct folding of the proteins during protein expression (38).
In conclusion, we here present evidence that the conserved hydrophobic amino acids in the C 128 -C 142 loop prevent ␣7-ECD from being water-soluble when expressed alone as a recombinant protein in heterologous expression systems. The replacement of this hydrophobic loop by the corresponding hydrophilic loop of the AChBP dramatically increases the solubility of the recombinant molecule and results in a molecule existing mainly in the form of an oligomer, probably a pentamer. This mutated recombinant human ECD is expressed in P. pastoris in satisfactory amounts and seems suitable as a model for structural studies of a functional LGIC, preferably after deglycosylation.