Water-soluble Nicotinic Acetylcholine Receptor Formed by α7 Subunit Extracellular Domains*

Water-soluble models of ligand-gated ion channels would be advantageous for structural studies. We investigated the suitability of three versions of the N-terminal extracellular domain (ECD) of the α7 subunit of the nicotinic acetylcholine receptor (AChR) family for this purpose by examining their ligand-binding and assembly properties. Two versions included the first transmembrane domain and were solubilized with detergent after expression inXenopus oocytes. The third was truncated before the first transmembrane domain and was soluble without detergent. For all three, their equilibrium binding affinities for α-bungarotoxin, nicotine, and acetylcholine, combined with their velocity sedimentation profiles, were consistent with the formation of native-like AChRs. These characteristics imply that the α7 ECD can form a water-soluble AChR that is a model of the ECD of the full-length α7 AChR.

AChRs are composed of five homologous membrane-spanning subunits that are ordered around a central, cation-selective channel. The topology of AChRs that is predicted by hydrophobicity plots has received substantial experimental support (4,5). The approximately 200 residues at the N-terminal half of each AChR subunit are extracellular, are N-glycosylated, contain sites for agonist and antagonist binding, and form the vestibule through which cations reach the transmembrane channel. Relatively little of the remainder of the primary sequence is extracellular. Three of the four transmembrane domains (M1-M3) that form the channel are grouped together following the N-terminal extracellular domain (ECD) and are separated from M4 by a large cytoplasmic loop. For the muscletype AChR, three distinct regions of the primary sequence around amino acid residues 86 -93, 149, and 190 -198 of ␣1 subunits and peptide loops around residues 34, 55-59, 113-119, and 174 -180 of the ␥ or ␦ subunit contribute to the ACh binding sites at the interfaces between ␣ and ␦ and between ␣ and ␥ (or ⑀) subunits, based on photoaffinity labeling and sitedirected mutagenesis (6,7). Because of primary sequence and topological similarity, homologous residues at subunit-subunit interfaces in other neuronal AChR subunits are expected to have similar roles in the agonist binding site. For example, residues of the ␣7 subunit homologous to those of ␣1 and to those around ␥55 and ␦57 have been shown to contribute to the agonist binding in homomeric ␣7 AChRs (8,9).
Our knowledge of the molecular structure of AChRs, however, is far from complete. Electron diffraction methods using two-dimensional, tubular arrays of AChRs from Torpedo californica have successfully yielded structural details at 9 Å resolution in three dimensions (10 -12) and at 7.5 Å in two-dimensional projection (13). Achieving higher resolution, however, has been elusive with membrane-bound or detergent-solubilized AChRs. No member of this superfamily of integral-membrane receptors has been crystallized, and the intact AChRs are too large (more than 200 kDa) for nuclear magnetic resonance spectroscopy.
An AChR formed from the ECD may be a suitable structural model of a full-length AChR, if the ECD can both fold and oligomerize in the absence of the remainder of the subunit. Several lines of evidence suggest that the ECD meets these requirements. First, the specific interactions between subunits that are important in assembly of the muscle-type AChR and for formation of mature acetylcholine binding sites appear to depend primarily on the ECD (14 -21). The long cytoplasmic loop of ␣1 participates in assembly subsequent to the formation of heterodimers (22). Second, membrane-tethered ECDs of mouse muscle ␣1 and ␦ form heterodimers with ligand binding sites that reflect properties of a full-length AChR (23)(24)(25). Third, sequences of ␣7 that affect homomeric assembly also have been localized to the first half of the ECD and an area around M1 based on chimeras of ␣7 and ␣3 (26). According to this report, the long cytoplasmic loop of ␣7 is not essential for oligomerization.
To determine whether AChRs formed from the ECD (residues 1-208) of ␣7 subunits are water-soluble structural models for the ECD of the full-length ␣7 AChR, we expressed two constructs of the ECD of the ␣7 subunit with M1 retained and one construct of the ECD without M1 in Xenopus oocytes. The constructs with M1 were included to explore the feasibility of removing M1 by in vitro processing subsequent to in vivo synthesis. We examined ligand binding properties and velocity sedimentation profiles as indicators of global structure and local structure at the agonist binding site of the resulting AChRs, which we have designated as ␣7 ECD AChRs. We found that each construct, including the water-soluble one without M1, assembles into an AChR with affinities for 125 Ilabeled ␣Bgt ( 125 I-␣Bgt), nicotine, and ACh that match those of the full-length ␣7 AChR. These properties demonstrate that the ␣7 ECD forms a water-soluble ␣7 ECD AChR that can be a starting point for structural studies of this superfamily of ion channels.

EXPERIMENTAL PROCEDURES
Design of ␣7M1, ␣7 Enterokinase (␣7EK), and ␣7 Water-soluble (␣7WS) Plasmids-The full-length cDNA sequence of chicken ␣7 (27) previously was cloned into a modified SP64T expression vector (28,29). For ␣7M1, which encodes the ECD of ␣7 up to the start of M2, the ␣7 coding sequence from the beginning of M2 to past the native stop codon was removed by digestion with BglII and BsmI. It was replaced by a double-stranded oligonucleotide cassette coding in-frame for the 19amino acid sequence SQVTGEVIFQTPLIKNPRV and a stop signal. This sequence contained the epitope of mAb 142 from residues 2 to 17 (5), followed by a MluI restriction site in the DNA sequence. The last native residue from ␣7 was Ile 240 , according to the numbering scheme of the mature chicken ␣7 AChR subunit (27). The mAb 142 epitope was introduced for immunoblotting and for binding the subunit protein to mAb 142-coated plastic wells for solid-phase assays (5). The MluI site was included so that the epitope insert could be extended with additional residues after digestion with MluI and BsmI. It has been shown that such extension may be necessary for accessibility of the epitope by the antibody (5).
The other construct that included M1 was ␣7EK, which was similar to ␣7M1 except for the inclusion of a peptide spacer between the end of the ECD and the beginning of M1. To prepare ␣7EK, a double-stranded oligonucleotide cassette coding for the 38-amino acid residue sequence TMRRRTGTVSISPESDRPDLSTFTSDDDDKILERRRTL was ligated in frame between the proximal BsmAI and distal HgaI sites that are nearly juxtaposed to the 5Ј side of M1 in ␣7M1. Residues 1-6 reconstructed the end of the N-terminal extracellular domain from residues Thr 203 to Thr 208 ; the DNA sequence for residues 7-8 introduced a KpnI site; residues 9 -23 were the epitope for mAb 236 (5); the DNA sequence for residues 24 -25 introduced a SpeI site; residues 26 -31 (DDDDKI) were the specificity sequence for EK that is modeled after its site of proteolysis on trypsinogen (30,31); the DNA sequence for residues 32-33 introduced a XhoI site; and residues 34 -38 reconstructed the native sequence of Arg 205 to Leu 209 at the distal end of the insert. Beginning with Tyr 210 (native numbering), the remainder of the sequence was identical to ␣7M1. The KpnI, XhoI, and SpeI restriction sites were included so that the mAb 236 epitope and the target for protease digestion easily could be modified readily. A total of 27 amino acid residues were inserted between the proximal copy of Thr 208 and the distal copy of Arg 205 .
The third construct, ␣7WS, was truncated at the end of the ECD of ␣7 and did not include M1. To prepare ␣7WS, the full-length sequence of ␣7 was cut at the BglII site between M1 and M2. The resulting Nterminal domain sequence was cut at the BsmAI site proximal to the start of the M1 coding sequence. A double-stranded oligonucleotide cassette coding for the 22-amino acid residue sequence TMRRRTQVT-GEVIFQTPLIKNP followed by a stop codon was ligated between the BsmAI site of the N-terminal sequence and the EcoRI site of a modified SP64T expression vector. Residues 1-6 of this sequence reconstructed the native sequence Arg 203 to Thr 208 ; residues 7-22 constituted the epitope for mAb 142. The oligonucleotides were synthesized using the phosphoramidite method on a MilliGen oligonucleotide DNA synthesizer.
Protein Expression in Xenopus oocytes-DNA from each of the three plasmids was purified on a CsCl gradient. cRNA was synthesized using an SP6 mMessage mMachine™ kit (Ambion, Austin, TX) and linearized DNA. The cytoplasm of each oocyte was injected with approximately 50 ng of cRNA and incubated at 18°C for 3-5 days in 50% Leibovitz's L-15 medium (Life Technologies, Inc.) in 10 mM HEPES, pH 7.5, containing 10 units/ml penicillin and 10 g/ml streptomycin.
For ␣7M1 and ␣7EK, extraction of membrane-bound subunit protein with a buffer containing Triton X-100 was accomplished with a previously reported procedure (29). Oocytes were homogenized by hand in ice-cold buffer A (50 mM sodium phosphate, 50 mM NaCl, 5 mM EDTA, 5 mM EGTA, 5 mM benzamidine, 15 mM iodoacetamide, pH 7.5). The membrane-containing fraction was separated by centrifugation, was washed twice with buffer A, and then was extracted with buffer B (40 mM sodium phosphate, 40 mM NaCl, 4 mM EDTA, 4 mM EGTA, 4 mM benzamidine, 12 mM iodoacetamide, 2% Triton) during gentle agitation for 2 hours at 4°C. The soluble fraction from this detergent extraction step was separated by centrifugation and was used for both Western blotting and assays of ligand binding.
For ␣7WS, the secreted fraction was defined as the subunit protein present in the L-15 medium incubating injected oocytes. Particulates in this fraction were removed by centrifugation. The cytoplasmic fraction of ␣7WS was defined as the subunit protein present in the soluble fraction following homogenization of the oocytes by hand in buffer A and centrifugation to sediment the insoluble, membranous component. The Triton-extracted fraction of ␣7WS was defined as the subunit protein present in the solvent following extraction in buffer B of the membranous component from the homogenization step during gentle agitation for 2 h at 4°C.
Immunoblotting-Triton-extracted ␣7M1 and ␣7EK or secreted ␣7WS was incubated overnight at 4°C with mAb 142 that had been coupled to Sepharose CL-4B (Pharmacia) with CNBr (32). After this concentration step, the protein was eluted at 55°C with 2% SDS and 20 mM ␤-mercaptoethanol. Proteins were deglycosylated for 18 h at 37°C with 1 unit of a mixture of endoglycosidase F and glycopeptidase F according to instructions of the manufacturer (Boehringer Mannheim). A sample without enzyme was run in parallel as the negative control.
Protein was denatured and reduced at 55°C in SDS-polyacrylamide gel electrophoresis sample buffer containing 2% SDS, separated on a 13% acrylamide SDS-polyacrylamide gel electrophoresis gel, and transferred to an Immobilon-P polyvinylidene difluoride membrane (Millipore). After being blocked in 5% powdered milk in phosphate-buffered saline (PBS, 100 mM NaCl, 10 mM sodium phosphate, pH 7.5) containing 0.5% Triton, the membrane was incubated with 2 nM 125 I-mAb 142. Specific activities of the labeled antibodies ranged from 10 17 to 10 18 cpm/mol. Labeling was visualized by autoradiography.
Ligand Affinities-Immulon 4 plastic microwells (Dynatech Laboratories, Chantilly, VA) were coated with mAb 142 or mAb 236 for solidphase assays (5). The wells were blocked with 3% bovine serum albumin in PBS. For ␣7M1 and ␣7EK, a volume of the Triton-solubilized protein from the equivalent of from one to three oocytes that had been injected with cRNA was added to each microwell. For the measurement of ␣Bgt affinity, 125 I-␣Bgt was added to the Triton extract and incubated overnight at 4°C. Total volume in each well was 100 l. The wells were washed three times with ice-cold PBS containing 0.5% Triton, and the amount of radioactivity was measured using a ␥ counter. Each data point was measured in duplicate. For the competitive inhibition assays, the wells were washed free of the Triton solution after 24 h and loaded with 125 I-␣Bgt at 4 nM in the presence of L-nicotine or ACh. The wells were incubated 8 h at 4°C before washing and then measuring the amount of radioactivity. Each data point was measured in duplicate. Nonspecific binding was measured with Triton-solubilized protein extracts from uninjected oocytes and generally was less than 5% of the specific binding.
For ␣7WS, the volume of L-15 medium above about six oocytes was added to mAb 142-coated microwells and incubated overnight at 4°C for capture of the secreted ␣7WS. The L-15 was washed away with PBS. For the measurement of ␣Bgt affinity in 0% Triton, 125 I-␣Bgt was added to buffer C (same composition as buffer B, except without Triton) and incubated overnight at 4°C. Total volume in each well was 100 l. The wells were washed three times with ice-cold PBS, and the amount of radioactivity was measured using a ␥ counter. Inhibition by nicotine and ACh in 0% Triton was measured after capture of the secreted ␣7WS, washing of L-15, and loading of each well with 0.6 nM 125 I-␣Bgt and the appropriate amount of inhibitor in buffer C. Total volume in each well was 100 l. The wells were incubated 8 h at 4°C before washing and measuring the amount of radioactivity. Data points were measured in duplicate. Nonspecific binding was measured with L-15 medium that was used to incubate uninjected oocytes and generally was less than 10% of the specific binding. Buffer B was substituted for buffer C at the step of loading the wells with 125 I-␣Bgt or with 125 I-␣Bgt and nicotine or ACh for the ligand affinity measurements of ␣7WS in 2% Triton.
The equilibrium dissociation constants K d for 125 I-␣Bgt were determined by least-squares, nonlinear fitting to a Hill-type equation (Equation 1) using the graphing software KaleidaGraph (Synergy Software) where C is the measured signal (counts/min), C 0 is the maximal signal (which corresponds in this case to the maximum number of 125 I-␣Bgt binding sites), L is the concentration of 125 I-␣Bgt, n is the Hill coefficient. The half-maximal inhibition constants, IC 50 , for nicotine and ACh in the presence of 125 I-␣Bgt were determined by nonlinear fitting to Equation 2, where L is the concentration of nicotine or ACh, C 0 is the maximal signal, and C 1 is a constant that represents signal that is not displaced by high concentrations of agonist. We used the Cheng-Prusoff equation (Equation 3) to estimate equilibrium dissociation constants from IC 50 values (33), although other equations also have been described for that purpose (34).
The K d values shown in Table I are the average and standard error of at least three independent assays of ligand affinity unless otherwise noted. Uncertainties shown in figures are standard errors. Sucrose Gradient Sedimentation-Membranes from 10 -20 oocytes that had been injected with the ␣7M1, ␣7EK, or full-length ␣7 cRNA were extracted in buffer B. AChRs in membrane vesicles from T. californica and AChRs from the TE671 cell line (35) were solubilized in buffer B. About 200-l aliquots of solubilized protein containing from 20 to 100 fmol of 125 I-␣Bgt binding sites were layered onto 5-ml sucrose gradients (5-20% (w/v)) in 0.5% Triton solution of 100 mM NaCl, 10 mM sodium phosphate, 5 mM EGTA, 5 mM EDTA, and 1 mM NaN 3 at pH 7.5. The gradients were centrifuged 75 min at 70,000 rpm (approximately 340,000 ϫ g) and 4°C in a Beckman NVT90 rotor. For determining a ligand binding profile, aliquots of 11 drops each (approximately 130 l) from the gradient were collected into Immulon 4 plastic microwells coated with mAb 142 for ␣7M1, mAb 236 for ␣7EK, or mAb 318 (a rat monoclonal antibody against an epitope in the cytoplasmic domain) (27) for the full-length ␣7. The entire gradient was collected in 40 fractions. After 24 h at 4°C, the microwells were washed and filled with from 4 to 12 nM 125 I-␣Bgt in buffer B for 6 h at 4°C, followed by washing and quantitation of bound 125 I-␣Bgt by ␥ counting.
Processing of ␣7WS was slightly more involved because of the low concentration of the secreted protein in the incubation medium. The protein was concentrated by binding overnight at 4°C to ␣-toxin that had been isolated from the venom of Naja naja siamensis with ion exchange chromatography (36) and then coupled to Sepharose CL-4B with CNBr (32). The bound protein was eluted from the ␣-toxin with 200 l of 100 mM nicotine and was layered onto 5-ml sucrose gradients (5-20%) without Triton in centrifuge tubes. The gradient was centrifuged 75 min at 70,000 rpm (approximately 340,000 ϫ g) and 4°C, as was done for ␣7M1 and ␣7EK. The entire gradient was collected in 39 fractions of 6 drops each, because the drop size was larger in the absence of detergent. After 24 h at 4°C, the microwells were washed and filled with 1 nM 125 I-␣Bgt in buffer C for 6 h at 4°C, followed by washing and quantitation of bound 125 I-␣Bgt by ␥ counting.
Comparison of AChR Yields-The yield was defined as the theoretical maximal amount of bound 125 I-␣Bgt, which was the value of C 0 in Equation 1 from 125 I-␣Bgt-binding assays. The AChR yield was calculated in terms of 125 I-␣Bgt binding sites per oocyte.

Expression of ␣7 ECD Proteins in Xenopus Oocytes-
The three variations of the ECD sequence of the ␣7 subunit were studied (Fig. 1). The first construct, ␣7M1, includes the Nterminal ECD, M1, and the portion of ␣7 between M1 and M2 up to residue Ile 240 . It was intended to demonstrate the pharmacological properties and oligomerization behavior of the ␣7 subunit proximal to M2 and tethered to the membrane through M1. The second construct, designated ␣7EK, contains a peptide spacer of 27 amino acid residues that is spliced between Thr 208 and Leu 209 at the junction of the ECD and M1. ␣7EK also contains M1 and terminates at Ile 240 , like ␣7M1. The position of an RRR motif from 205 to 207 at the junction of the ECD with M1 suggests a structural role for these positively-charge residues; therefore, the native sequence Arg 205 to Thr 208 was repeated at the C-terminal end of the interposed segment before M1. The name "␣7EK" is derived from the DDDDKI sequence that was included in the peptide spacer for site-specific proteolysis by EK. ␣7EK was intended to test the feasibility of removing M1 from the ECD by in vitro enzymatic proteolysis after synthesis. The third variation, ␣7WS, was intended to be a direct route to a water-soluble ␣7 ECD AChR. It is truncated at Thr 208 , just before M1 and contains no transmembrane domain. Epitopes for monoclonal antibodies mAb 142 and mAb 236 were included in the truncated constructs so that the proteins could be detected by immunoblotting and tethered through the antibodies to plastic wells for solid-phase ligandbinding assays (5).
Immunoblotting of Triton X-100 extracts from oocytes injected with ␣7M1 and ␣7EK cRNA confirmed the expression of these proteins (Fig. 2). ␣7M1 before deglycosylation migrated at an apparent mass of 37 kDa; ␣7EK migrated at 40 kDa. These apparent masses were about 7 kDa higher than the values of 30 kDa for ␣7M1 and 33.5 kDa for ␣7EK that were FIG. 1. Designs of the three ␣7 ECD subunit proteins ␣7M1, ␣7EK, and ␣7WS. The ECD, which extends from residues 1 to 208 of the mature chicken ␣7 AChR (27) is included in each protein. In ␣7M1 and ␣7EK, the first transmembrane domain, M1, extends from residues 209 to 233. The ␣7 sequence from residues 234 to 240, which extends to the beginning of M2, is included after M1 in these two proteins. The segment labeled 142 t on the C-terminal side of M1 is the epitope tag for mAb 142. Besides the mAb 142 epitope, ␣7EK also contains an epitope tag for mAb 236, designated 236 t , as well as an EK-specific protease site between the extracellular domain and M1. In contrast to the two membrane-tethered proteins, ␣7WS does not contain M1; the mAb 142 epitope tag of the ␣7WS protein follows directly after the ECD. calculated from amino acid compositions. The detection of predominately a single band before deglycosylation suggested that post-translational modifications were comparatively uniform on all molecules. Deglycosylation shifted each band to approximately the molecular mass of the protein calculated without modifications.
Without M1, ␣7WS was secreted in soluble form into the incubation medium by oocytes that had been injected with ␣7WS cRNA. The secreted protein displayed an apparent mass of about 41 kDa by immunoblotting, compared with a calculated mass of 26 kDa (Fig. 2). The detection of a diffuse pattern of bands before deglycosylation suggested more heterogeneous glycosylation than was present on ␣7M1 and ␣7EK. Deglycosylation shifted the apparent mass down to about 26 kDa, confirming that ␣7WS had the most extensive glycosylation of the three proteins. The larger shift in apparent mass compared with ␣7M1 and ␣7EK also suggested that the pattern of glycosylation differed between membrane-tethered and soluble forms of the ECD. An additional observation from the immunoblot is that the yield of secreted ␣7WS was about 10-fold less than the yields of ␣7M1 and ␣7EK from Triton-solubilized membrane fractions, based on a comparison of the numbers of injected oocytes needed to produce approximately equal signals on the immunoblot.
High Affinity Binding of 125 I-␣Bgt to ␣7 ECD AChRs-Because high affinity binding of ␣Bgt is a hallmark of the ␣7 AChR, measurements of binding affinity with 125 I-␣Bgt were appropriate starting points for analyzing the functional and structural properties of the three ␣7 ECD constructs. The equilibrium dissociation constant, K d , of 125 I-␣Bgt was 1.6 nM for ␣7M1 and 1.9 nM for ␣7EK when measured in solid phase assays in which the Triton-solubilized proteins were bound in mAb 142-coated wells ( Fig. 3 and Table I). These values compare favorably with the published K d of 1.6 nM for 125 I-␣Bgt binding to Triton-solubilized full-length ␣7 AChR and for ␣7containing AChR from chick brain (37). Binding of 125 I-␣Bgt was measured for ␣7EK by also using the mAb 236 epitope tag to bind the protein to mAb 236-coated wells. The K d was 2.3 nM, demonstrating that the mAb 236 epitope tag at the C-terminal end of the ECD did not interfere significantly with the function of the ␣Bgt binding site.
The affinity of secreted ␣7WS for 125 I-␣Bgt was measured first without Triton (Fig. 4 and Table I). The K d value of 0.4 nM was about 4-fold smaller than the values for Triton-solubilized ␣7M1, ␣7EK, and full-length ␣7. The presence of 2% Triton during the binding assay of ␣7WS for 125 I-␣Bgt increased the value of K d to 1.7 Ϯ 0.1 nM (Fig. 4), which is in the range measured for the Triton-solubilized AChRs. Triton, however, did not significantly affect the total number of 125 I-␣Bgt binding sites. The ratio of the number of binding sites calculated from Equation 1 in the absence of Triton compared with the number of binding sites in the presence of Triton from otherwise identical pools of secreted ␣7WS was 0.99 Ϯ 0.13 (n ϭ 2). Only the secreted fraction of ␣7WS bound 125 I-␣Bgt to a measurable extent, despite the significant amount of ␣7WS protein that was retained intracellularly. The total amount of ␣7WS protein per oocyte that was detected in the cytoplasmic and the Triton-solubilized membrane fractions by immunoblotting was approximately equal to the amount of Triton-solubilized ␣7M1 or ␣7EK protein in similarly-injected oocytes. Hence all three constructs apparently were synthesized and accumulated to an approximately equal extent. The intracellular pool of ␣7WS, however, showed no significant binding of 125 I-␣Bgt.
The yield of 125 I-␣Bgt binding sites from each of the three ECD proteins was measured by the binding of 125 I-␣Bgt to AChR on mAb 142-coated wells (Fig. 5). These measurements assumed that the mAb 142 epitope in each protein was equally accessible to the antibody. The yield of 8 Ϯ 1 fmol of 125 I-␣Bgt binding sites per oocyte with ␣7M1 AChR was the largest among the three proteins. This value is near the value of 15 fmol per oocyte that was reported for Triton-solubilized AChR from the full-length ␣7 subunit expressed in oocytes (29). Compared with ␣7M1, the addition of the peptide spacer in ␣7EK decreased the yield of AChR by about three-quarters to 1.9 Ϯ 0.6 fmol of 125 I-␣Bgt binding sites per oocyte. The yield of secreted ␣7WS AChR was the smallest of the three at 0.21 Ϯ 0.02 fmol of 125 I-␣Bgt binding sites per oocyte. These comparisons indicate that insertion of residues before M1 or the absence of M1 decreases the yield of ECD AChRs.
High affinity for 125 I-␣Bgt was the first step in demonstrating the structural similarity of ␣7M1, ␣7EK, and ␣7WS AChRs to the full-length ␣7 AChR. It was only an initial test of the properties that we hoped to find in a model AChR. Specifically, we were seeking a model that also was a pentamer and that had native-like affinity for agonists like nicotine and ACh. High affinity for ␣Bgt does not necessarily imply either of these properties. Muscle-type AChR illustrates this point. The human ␣1 subunit in the absence of other muscle-type subunits binds ␣Bgt relatively tightly, with K d of 0.6 nM compared with K d of the pentameric muscle-type AChR of 0.1 nM (35). Instead of a pentamer, however, it exists as a monomer according to its velocity sedimentation profile relative to the profiles of fully assembled muscle-type AChR and ␣␦ and ␣␥ dimers (38,39). Moreover, it does not bind agonists with native-like affinity. In light of these characteristics for ␣1, other experiments besides binding of 125 I-␣Bgt were required for assessing the oligomerization of the ␣7 ECD constructs.
Velocity Sedimentation Implies ␣7M1, ␣7EK, and ␣7WS Form Multimeric AChRs, Probably Pentamers-Oligomerization of ␣7M1, ␣7EK, and ␣7WS proteins was examined first with velocity sedimentation profiles in sucrose gradients. The proteins were detected by binding of 125 I-␣Bgt in solid-phase assays. These profiles were compared with full-length AChRs and monomeric ␣1 subunits as sedimentation standards. Fully assembled, full-length ␣7 AChR (calculated molecular mass of 272 kDa from amino acid sequence) were pentameric standards (37). Muscle-type AChR from T. californica were pentameric (i.e. a monomeric AChR with calculated molecular mass of 268 kDa) and decameric standards (i.e. a dimeric AChR with calculated molecular mass 536 kDa) (40). The monomeric ␣1 subunit (calculated molecular mass of 52 kDa) from the human rhabdomyosarcoma cell line TE671 was the monomeric standard (35). As expected because of their much larger molecular masses, the pentameric AChR standards sedimented faster than monomeric ␣1 subunit (Fig. 6).
The single 125 I-␣Bgt binding peak from ␣7M1 and from ␣7EK also sedimented much faster than the monomeric ␣1 subunit and slightly slower than the full-length AChRs (Fig. 6). Two conclusions can be drawn from these velocity sedimentation data in the absence of measurements of diffusion and partial specific volume. First, the position of the 125 I-␣Bgt binding peak was distant from the ␣1 monomer standard, strongly suggesting that a multimeric species was responsible for the 125 I-␣Bgt binding. Second, this multimeric species probably was a pentamer. This conclusion about subunit stoichiometry was deduced from both the single peak and its relatively narrow distribution. The single peak implied that a multimer with a single distinct stoichiometry was the dominant species that bound 125 I-␣Bgt with nanomolar affinity. In contrast, if multiple stoichiometries such as dimers and trimers and tetramers had significant representation in the high affinity population, then one would have expected either several discrete peaks or a low, broad profile. Because of the pentameric structure of the full-length ␣7 AChR (41), the most likely stoichiometry for ␣7M1 and ␣7EK AChRs also was pentameric, although minor fractions of smaller oligomers may also have been present. Moreover, it has been reported that only pentameric ␣7 subunits bind 125 I-␣Bgt with nanomolar affinity, whereas monomeric full-length subunits and aggregated or incompletely assembled forms do not (37).
Velocity sedimentation of the secreted ␣7WS in sucrose gradients also revealed a single 125 I-␣Bgt binding peak (Fig. 6). It sedimented slightly slower than the peaks for ␣7M1 and ␣7EK. As with ␣7M1 and ␣7EK, the position of the single, relatively narrow peak that was distant from the peak of the ␣1 subunit did not suggest a monomeric form for ␣7WS AChR and was consistent with the conclusion that the dominant form of ␣7WS AChR was a pentamer.
High Affinity Binding of Nicotine and ACh to ␣7 ECD AChRs-High affinity binding for agonists was the second test of oligomerization and the last test of structural equivalence between the ␣7 ECD AChRs and the full-length ␣7 AChR. Interpretation of this test was based on an analogy with characteristics of the ␣1 subunit of the muscle-type AChR. High affinity for agonists develops for ␣1 only after it assembles with ␦ or ␥ (39,42), because the agonist binding site consists of structural components from both subunits at ␣␦ and ␣␥ interfaces (3). Similarly, if ␣7M1, ␣7EK, and ␣7WS AChRs indeed were pentamers and structural models of the full-length ␣7 AChR, then we expected native-like affinities for nicotine and ACh.
The affinities of ␣7M1 and ␣7EK for the small ligands nicotine and ACh were measured by competitive inhibition of the binding of 125 I-␣Bgt ( Fig. 7 and Table I). The equilibrium dissociation constants, K d , of ␣7M1 and ␣7EK were 1.0 M for nicotine and 50 M for ACh. In the case of ␣7EK, tethering in the solid phase assays via the mAb236 epitope tag slightly increased the values of K d for nicotine and ACh to 1.4 M and 60 M, respectively. For both proteins, the K d values are similar to those for Triton-solubilized, full-length ␣7 and for ␣7-containing AChRs from chick brain (Table I). Competitive binding with each ligand eliminated almost all binding of 125 I-␣Bgt, implying that the majority of the 125 I-␣Bgt binding sites also bound nicotine and ACh.
In inhibition assays with 125 I-␣Bgt and ␣7WS, the values of K d for nicotine binding in the absence of Triton were 10-fold smaller than the values for nicotine binding to ␣7M1 and ␣7EK in the presence of Triton ( Fig. 8 and Table I). The values for ACh binding to ␣7WS without Triton were 40-fold smaller than the values for ACh binding to ␣7M1 and ␣7EK with Triton. The water-soluble ␣7WS AChR without Triton, therefore, bound to small ligands more tightly than did the full-length ␣7 AChR in Triton. The presence of 2% Triton in the inhibition assays increased the values of K d by about 3-fold (Fig. 8). As with ␣7M1 and ␣7EK, the majority of the 125 I-␣Bgt binding sites of ␣7WS AChR also bound nicotine and ACh, as shown by the elimination of almost all binding of 125 I-␣Bgt at high concentrations of agonists in either the presence or absence of Triton.
In light of the velocity sedimentation data, these results confirm that the ␣7M1, ␣7EK, and ␣7WS AChRs are oliogomers and probably are pentamers. These results demonstrate that the ␣7 ECD with or without M1 is sufficient for the expression of a ␣7 ECD AChR with native-like affinities for ␣Bgt and nicotinic agonists.

DISCUSSION
Structural Implications of the Ligand Binding Properties and Oligomerization of ECD AChRs-We have shown that the N-terminal ECD of the ␣7 AChR subunit with or without M1 forms an AChR whose ligand-binding and sedimentation prop-  Table I. In the absence of Triton, K d was about 4-fold smaller than the values measured for ␣7M1 and ␣7EK. In the presence of Triton, however, the value for ␣7WS shifted into the range for the M1-containing AChRs. Data from representative single experiments are shown.

FIG. 5.
Comparison of yields of ␣7 ECD AChRs from ␣7M1, ␣7EK, and ␣7WS. The yield was defined as the number of 125 I-␣Bgt binding sites per oocyte. This value was the maximal amount of bound 125 I-␣Bgt that was estimated from solid phase binding assays on mAb 142-coated microwells using Equation 1 (calculated from the value of C 0 ). At least three independent measurements were performed for each type of AChR. From 20 to 50 oocytes were pooled for the assays of ␣7M1 and ␣7EK. The incubation medium from 80 to 160 oocytes was the source of AChRs for the assays of ␣7WS. erties imply that it is a structural model of the full-length AChR. In particular, the ␣7WS AChR is water-soluble. The ECD, therefore, constitutes an autonomous unit capable of folding and assembling in the absence of the transmembrane, cytoplasmic, and other extracellular portions of the ␣7 subunit sequence. We examined three characteristics that implied retention of the native-like structure in the three ECD AChRs: 1) high affinity binding of ␣Bgt, which is a hallmark of ␣7 AChRs; 2) velocity sedimentation, which is most consistent with assembly of each of the three ␣7 ECD subunits into pentamers; and 3) native-like affinities for nicotine and ACh.
First, the affinity for 125 I-␣Bgt of each of the three ␣7 ECD AChRs was equal to or higher than that of the fully assembled, full-length ␣7 AChR. Because the ␣Bgt binding site incorporates noncontiguous regions of primary sequence, these affini-ties suggest that the global structure of each ␣7 ECD AChR models the structure of the ECD in the full-length ␣7 AChR. The moderately higher affinity of the ␣7WS AChR for 125 I-␣Bgt reverted back into the range of K d values observed with ␣7M1 and ␣7EK AChRs when Triton was included in the assay. Triton could have exerted its effects by modifying the interaction of ␣Bgt with either the AChR or with the solvent environment, since both interactions can affect K d . In other words, Triton may have caused structural perturbations of the AChR that affected the binding site, or 125 I-␣Bgt may have partitioned differently in Triton-containing solvent compared with detergent-free solvent.
Second, velocity sedimentation data suggested that the ␣7 ECD subunits had assembled into multimers that probably were pentamers. Other AChRs or AChR subunits were used as oligomerization standards for estimating the extent of oligomerization of the ␣7 ECD subunits. The single 125 I-␣Bgt binding peak for each of the ECD AChRs migrated on the sucrose gradient slightly slower than full-length, pentameric ␣7 AChR and Torpedo AChR and much faster than the peak of monomeric ␣1 subunit. Although differences in the amount of FIG. 6. Velocity sedimentation profiles of ␣7 ECD AChRs and full-length ␣7 AChRs. The profiles of 125 I-␣Bgt binding to full-length ␣7 AChR, ␣7M1 AChR, and ␣7EK AChR that were sedimented on a 5-20% sucrose gradient with 0.5% Triton X-100 are shown in the upper panel. Two of the arrows indicate the peaks of 125 I-␣Bgt binding to Torpedo AChR monomers (five subunits) and dimers (10 subunits) (40) that were observed in a parallel experiment. The peak of 125 I-␣Bgt binding to the human ␣1 monomeric subunit from the TE671 cell line from a parallel experiment is indicated by a third arrow. These three proteins served as oligomerization standards for the ␣7 ECD AChRs. The profiles for ␣7M1 and ␣7EK are inconsistent with monomers as the high affinity species but are consistent with pentamers. The lower panel shows the profile of the ␣7WS AChR on a 5-20% sucrose gradient without Triton. The profile again is inconsistent with a monomer as the high affinity species but is consistent with a pentamer.  Table I. Similar inhibition assays were performed with ␣7EK using mAb 236-coated microwells (Table I). Data from representative single experiments are shown.
bound detergent (40) and in molecular shape (43,44) will affect the sedimentation of AChRs, we do not think that these effects alone can account for the differences between the position of the monomeric subunit standard and the positions of ␣7 ECD AChRs on the gradients. A reasonable interpretation of the position of the 125 I-␣Bgt binding peaks relative to the monomeric ␣1 subunit peak was that these ␣7 ECD subunits were not monomeric. Multimeric stoichiometry also is consistent with their native-like affinity for agonists. Given that fulllength ␣7 AChR are pentamers (41), the single, relatively narrow 125 I-␣Bgt binding peak from each ␣7 ECD AChR probably arose from pentamers.
Third, affinities of the three ␣7 ECD AChRs for nicotine and ACh also matched or were slightly higher than those of the full-length ␣7 AChR. In the case of ␣7WS AChR, Triton reduced the affinity of nicotine and ACh compared with that observed without Triton. The differences in affinity for agonists of ␣7WS AChR compared with ␣7M1 and ␣7EK AChRs were not caused solely by Triton and may have been contributed in part by M1. Overall, these results imply that the local structure of the ACh binding site of each of the ␣7 ECD AChRs closely matches that of the full-length ␣7 AChR. By analogy with the behavior of the muscle-type AChR (39,42), the binding site for small ligands is thought to form at interfaces between subunits. Therefore, native-like affinities for agonists by the ␣7 ECD AChRs, combined with the results from velocity sedimentation, confirmed that the ␣7 ECD subunits had assembled into multimers.
The ␣7EK subunit demonstrated another structural property of the ECD AChRs. Splicing a peptide spacer of 27 amino acid residues between these two domains did not significantly alter ligand affinity. Therefore, strict continuity of the native primary amino acid sequence between the ECD and M1 is not required for formation of the agonist binding site or for oligomerization. In other words, an ECD that is displaced away from M1 by a peptide spacer retains its assembly and binding properties. This flexibility in design raises the possibility of first expressing a membrane-bound ECD AChR, followed by release of a water-soluble ECD AChR from M1 by enzymatic proteolysis in vitro.
Elimination of transmembrane domains either by recombinant techniques or by proteolysis of native protein has been the starting point for structural studies of other integral membrane proteins. In the family of ionotropic glutamate receptors, the soluble agonist binding domain of GluR-B and GluR-D was successfully designed by fusion of discontinuous sections of the primary sequence that apparently are separated by two transmembrane domains (45,46). This strategy also has been successful for the x-ray crystallography of many integral membrane proteins including growth hormone receptor (47), prolactin receptor (48), tissue factor (49), interferon-␥ receptor (50), human class II histocompatibility antigen (51), insulin receptor protein-tyrosine kinase domain (52), ␣ and ␤ chains of the T cell receptor (53,54), neuraminidase (55), hemagglutinin (56), and bacterial aspartate receptor (57).
Yield of ECD AChRs-The protein sequence between the beginning of M1 and the start of M2 (residues Leu 209 to Ile 240 ) was needed for production of AChRs from the ECD at a level comparable to that of the full-length AChR. M1 probably was the key component of this region. Compared with the yield of AChRs from ␣7M1 AChR, the yield was smaller from ␣7EK and even smaller from ␣7WS. The differences were unlikely to be caused by reduced translation efficiencies of ␣7EK and ␣7WS, because of the extent of sequence identity among all three designs and because the amounts ␣7EK and intracellular ␣7WS proteins per oocyte that were detected by immunoblotting were approximately equal to the amount of ␣7M1. Instead, the difference in the case of ␣7EK AChR suggests some adverse effects on folding and assembly caused by the alteration of the primary sequence at the ECD/membrane interface.
The low yield in the case of ␣7WS AChR highlighted an important role for M1. M1 has been shown to function as a tether to the ER membrane during the dimerization of truncated ␣1 and ␦ subunits in the ER (23). It was successfully replaced in mouse ␣1 by unrelated transmembrane domains with unrelated primary sequences or with a glycosylphosphatidylinositol (GPI) moiety (24). A similar role for M1 probably applies to the production of ␣7 ECD AChRs. M1 constrains the subunits to a membrane surface, which may favor orientations between subunits that are required for assembly and may help retain them in the endoplasmic reticulum (ER) and Golgi apparatus for folding, subunit assembly, and post-translational modifications. In contrast, failure to retain the subunits on the ER and Golgi membranes removes such constraints and decreases the efficiency of these processes. In particular, M1 may enhance folding and assembly by increasing the local concentration of subunits relative to folding cofactors, such as calnexin, that are found in the ER membrane (58,59).
Attaching the ECD to a membrane may not be the sole function of M1 for ␣7M1 and ␣7EK, however. Instead, the yield of AChRs may depend on particular properties of M1. Chimeras of ␣7/␣3 (26) and point mutations in M1 of ␣7 (60) demonstrate that the specific sequence used in the first transmembrane domain affects the yield of surface AChRs. Similarly, the total amount of bound 125 I-␣Bgt per oocyte varied when unrelated transmembrane domains replaced any one of the regions M1-M4 of ␣1 in recombinantly expressed Torpedo AChRs (61).
Strategy for Structural Studies of ␣7 ECD AChRs-What is the optimum design of an ␣7 ECD AChR for structural studies? The complete ECD of the ␣7 subunit incorporating all of the amino acid sequence up to the start of M1 probably is necessary to satisfactorily reproduce the ligand-binding and assembly properties of the full-length ␣7 AChR. Although we did not attempt to truncate the ␣7 ECD N-terminal to Thr 208 , results from the mouse ␣1 subunit suggest that there is little flexibility for moving the truncation point closer to the N terminus and into residues known to contribute to the ␣Bgt binding site (23). Truncation of mouse ␣1 after Pro 211 , which is homologous with our truncation point in ␣7WS, did not disrupt the formation of a high affinity binding site for ␣Bgt. In contrast, truncation of the mouse ␣1 subunit after Met 207 (mouse ␣1 numbering) caused a loss of affinity for ␣Bgt.
More extreme truncations of the ECD of ␣1 have been explored in attempts to bypass the difficulties presented by AChRs for structural studies. None, however, appears as successful in duplicating properties of full-length AChRs as the ␣7 ECD AChRs described here. For example, peptide sequences from the region around residues 170 -200 of the ␣1 subunit have been used as potential mimics of the ␣Bgt binding site for NMR and ligand-binding studies (62,63). The affinity for the binding of ␣Bgt, however, typically is three orders of magnitude less than with full-length AChRs (64). In addition, peptide models from the ␣1 subunit are incapable of oligomerization and do not show significant affinity for small ligands.
Three conclusions from our investigation are relevant to a strategy for obtaining ␣7 ECD AChRs for structural studies. First, the extracellular domain constitutes a stable, fully water-soluble AChR that quantitatively retains essential ligandbinding properties of the full-length AChR and eliminates confounding factors of detergent solubilization that hamper the crystallography of membrane proteins. Second, M1 in the ␣7M1 and ␣7EK constructs significantly enhances the yield of ECD AChRs compared with the yield from ␣7WS. Although membrane tethering of the AChR subunit proteins was essential for association of the ECD of the mouse muscle-type ␣1 subunit and the full-length ␦ subunit when expressed in COS cells (24), it was advantageous but not essential for assembly of ␣7WS AChR in oocytes. Third, the ␣7 ECD still can form an AChR even when a peptide spacer separates it from M1 and the adjacent membrane surface.
These findings suggest that a water-soluble ␣7 ECD AChR more likely will be produced in large amounts in two stages via a membrane-bound intermediate than directly from a watersoluble design of the ECD. The first stage is in vivo synthesis of a membrane-bound AChR that subsequently will be removed from its membrane tether by in vitro processing. The second stage is enzymatic removal of the membrane tether in vitro. ␣7EK demonstrates that an AChR substrate can be designed for specific proteolysis within a peptide spacer between the ECD and M1. Our attempts to separate an ECD AChR from the M1 domain of ␣7EK using EK, however, resulted in substantial nonspecific proteolysis. An alternative method that allows truncated ␣1 subunits to dimerize with full-length ␦ is substitution of M1 with a GPI moiety (24,65). A GPI tether also leads to the expression of membrane-bound extracellular domains of chick ␣7 on the surface of oocytes (66). A GPI tether may be preferred at the stage of enzymatic processing, because the carbohydrate portion of the anchor at the C terminus of the protein can be separated from the more distal, membraneembedded fatty acid portion with phosphatidylinositol-specific phospholipase C (67,68). This enzyme is expected to eliminate the risk of nonspecific enzymatic proteolysis.
In addition to ␣7, we anticipate that ECD receptors of other members of the family of AChR subunits and other subunits of this entire superfamily of neurotransmitter-gated ion channels will mimic the structure and function of their respective fulllength receptors. Demonstrations that ECD sequences are important in the assembly of both glycine receptors (69) and ␥-aminobutyric acid A receptors (70) suggest that the ECD of other members of the superfamily also will fold and assemble autonomously. The ultimately successful design of an ECD AChR for high level expression may require other design modifications. Moreover, structural aspects of the gating and ion permeation functions of transmembrane domains will have to be explored with different strategies. A water-soluble, recombinantly generated ECD AChR, however, appears to be a promising foundation for structural studies of this superfamily of ion channels.