Extracellular Domain Nicotinic Acetylcholine Receptors Formed by α4 and β2 Subunits*

Models of the extracellular ligand-binding domain of nicotinic acetylcholine receptors (nAChRs), which are pentameric integral membrane proteins, are attractive for structural studies because they potentially are water-soluble and better candidates for x-ray crystallography and because their smaller size is more amenable for NMR spectroscopy. The complete N-terminal extracellular domain is a promising foundation for such models, based on previous studies of α7 and muscle-type subunits. Specific design requirements leading to high structural fidelity between extracellular domain nAChRs and full-length nAChRs, however, are not well understood. To study these requirements in heteromeric nAChRs, the extracellular domains of α4 and β2 subunits with or without the first transmembrane domain (M1) were expressed in Xenopus oocytes and compared with α4β2 nAChRs based on ligand binding and subunit assembly properties. Ligand affinities of detergent-solubilized, extracellular domain α4β2 nAChRs formed from subunits with M1 were nearly identical to affinities of α4β2 nAChRs when measured with [3H]epibatidine, cytisine, nicotine, and acetylcholine. Velocity sedimentation suggested that these extracellular domain nAChRs predominantly formed pentamers. The yield of these extracellular domain nAChRs was about half the yield of α4β2 nAChRs. In contrast, [3H]epibatidine binding was not detected from the extracellular domain α4 and β2 subunits without M1, implying no detectable expression of extracellular domain nAChRs from these subunits. These results suggest that M1 domains on both α4 and β2 play an important role for efficient expression of extracellular domain α4β2 nAChRs that are high fidelity structural models of full-length α4β2 nAChRs.

Nicotinic acetylcholine receptors (nAChRs) 2 are ligand-gated ion channels expressed mainly in the nervous system and at the neuromuscular junction (1)(2)(3), although they also are expressed elsewhere (4). They are members of the superfamily of nicotinoid receptors, which includes ␥-aminobutyric acid (GABA) type A and C, glycine, and serotonin 5-HT3 receptors. They contain five homologous subunits, with subunits designated ␣1-␣10, ␤1-␤4, ␦, ␥, and ⑀. Each subunit is an integral membrane protein with a large, ligand-binding N-terminal extracellular domain, four transmembrane domains (M1-M4), and a cytoplasmic loop between M3 and M4. The study of structure and dynamics of nAChRs has been motivated by their roles in normal and pathologic neurophysiology, including roles in addiction, neurodegeneration, epilepsy, myasthenia gravis, and congenital myasthenic syndromes (5).
Knowledge of the structure of nAChRs has come from biochemical, electrophysiological, and imaging methods (6 -8). These methods, however, do not provide comprehensive structure at atomic resolution. A significant advance toward such information has come recently from the crystallographic structure of acetylcholine-binding protein (AChBP), a water-soluble pentamer that has been the basis for homology modeling of the extracellular domain of nAChRs and other nicotinoid receptors (9 -18). Continuing to move to structures of native nAChRs is needed for accurate knowledge of how these receptors function as ion channels and signal transducers, for understanding differences in pharmacological profiles and kinetics of different receptors, and for designing drugs to therapeutically target specific types of receptors for specific effects in the central and peripheral nervous systems. Structures of full-length nAChRs have been elusive because they are large, integral membrane proteins.
One attractive strategy to make nAChRs more tractable for high resolution structural analysis is to produce water-soluble models from the extracellular domain without transmembrane domains. Compared with full-length nAChRs, water-soluble extracellular domain nAChRs are expected to crystallize more readily and, because of their smaller molecular mass, are expected to be better candidates for structural studies by NMR spectroscopy. This approach has shown promise for nAChRs. Models of extracellular domains of nAChRs have ranged from short amino acid sequences around the agonist-binding site (19 -23) to most or all the extracellular domain from ␣1 or ␣7 subunits expressed in bacteria or yeast (24 -30). These models, however, have only partially reproduced ligand binding affinities and subunit assembly properties of full-length receptors, motivating a search for models that more fully reflect the properties of full-length receptors.
Extracellular domain nAChRs are defined here as comprising truncated subunits containing N-terminal extracellular domain sequences and having ligand binding affinities or subunit assembly properties similar to full-length nAChRs. They have been expressed with several subunit combinations. Dimeric extracellular domain nAChRs were produced by extracellular domains of ␣1 and ␦ subunits that also contained M1 or another membrane tether, suggesting that membrane attachment of subunits might be important for forming these extracellular domain nAChRs (31,32). Extracellular domains of ␣7 subunits linked to a glycosylphosphatidylinositol moiety were transported to the surface of Xenopus oocytes and bound ␣-bungarotoxin (␣Bgt) and d-tubocurarine (33), implying the formation of multimers. Water-soluble extracellular domain ␣7 nAChRs without any transmembrane domains bound ␣Bgt, nicotine, and acetylcholine with affinities similar to ␣7 nAChRs and likely were pentamers (34). Their yield, however, was substantially smaller than the yield of detergent-solubilized, extracellular domain ␣7 nAChRs that contained M1, suggesting that M1 was important for expressing these extracellular domain nAChRs. Milligram amounts of a water-soluble, extracellular domain chimera containing ␣7 and AChBP sequences were produced in yeast (26). These extracellular domains formed oligomers that likely were pentamers. Water-soluble, pentameric, and extracellular domain muscle-type receptors without transmembrane domains have been demonstrated with electron microscopy (35).
Different subunit combinations might more readily produce pentameric, water-soluble nAChRs than do extracellular domain ␣7 or muscletype subunits. To extend the concept of extracellular domain nAChRs to other subunits and to investigate further the role of M1 in the formation of extracellular domain nAChRs, ␣4 and ␤2 subunits were studied because they produce heteromeric nAChRs and must form at least dimers to produce high affinity binding sites and because they offer potential for high level expression (36). Extracellular domain ␣4 and ␤2 subunits with M1 produced nAChRs in Xenopus oocytes with ligand affinities nearly identical with those of ␣4␤2 nAChRs. The subunit stoichiometry of extracellular domain ␣4␤2 nAChRs with M1 likely was pentameric. Without M1, however, no extracellular domain nAChRs were detected, suggesting that M1 plays an important role in the efficient expression of extracellular ␣4␤2 nAChRs.

EXPERIMENTAL PROCEDURES
Design of Subunit DNA Plasmids-The cDNAs for human ␣4 subunit (37) and human ␤2 subunit (38) were cloned into pSP64 poly(A) (Promega). The DNA sequences for ␣4M1, ␤2M1, ␣4WS, and ␤2WS extracellular domain constructs ( Fig. 1) were derived from the ␣4 or ␤2 sequences by PCR. The N-terminal primers for PCR were GCGC AAGCTT ATG GAG CTA GGG GGC CCC GGA for both ␣4-derived extracellular domain constructs and GCGC CTGCAG ATG GCC CGG CGC TGC GGC CCC for both ␤2-derived extracellular domain constructs. The C-terminal primers for PCR were GCGC ACGCGT CTC GGA GGG CAG GTA GAA GAC for ␣4M1, GCGC ACGCGT GTC GGA TGG CAG GTA GAA GAC for ␤2M1, GCGC ACGCGT CAG CCG CCG GAT GAC GAA GGC for ␣4WS, and GCGC ACGCGT CTT GCG GCG AAT GAT GAA GTC for ␤2WS. PCR with Pfu DNA polymerase (Stratagene) was performed with 10 ng of template DNA and 50 pmol of each primer for 30 cycles in 4% Me 2 SO with annealing at 60°C for 1 min and extension at 72°C for 3 min. The ␣4M1 and ␣4WS products from PCR were cut with HindIII and MluI; the ␤2M1 and ␤2WS products were cut with PstI and MluI. These restricted PCR products were ligated into pSP64 poly(A) along with a double-stranded oligonucleotide cassette coding in-frame for two copies of the mAb 142 epitope QVTGEVIFQTPLIKNP (39,40) for ␣4M1, one copy of the mAb 142 epitope for ␣4WS, and one copy of the mAb 236 epitope VSISPESDRPDLSTF (39,40) for ␤2M1 and ␤2WS. The DNA sequences of each epitope tag contained an MluI site before and a stop codon after the sequence coding the epitope tag. To reduce the length of subunit sequence that was derived from PCR, the ␣4 sequence upstream from its BstYI restriction site and the ␤2 sequence upstream from its BsiWI were replaced by DNA sequences that were taken from the full-length subunits. The last native residues in the constructs were Glu-270 (␣4 numbering from Swiss-Prot (41) entry P43681) for ␣4M1, Leu-242 for ␣4WS, Asp-261 (␤2 numbering from Swiss-Prot entry P17787) for ␤2M1, and Lys-233 for ␤2WS. The DNA sequences for ␣4M1 NT and ␤2M1 NT were prepared by replacing the MluI site at the 5Ј end of the epitope tag coding sequence with an MluI site followed by a stop codon. The DNA sequences for ␣4WS NT and ␤2WS NT were prepared by replacing the native coding sequence distal to the BglII site near the 3Ј end of ␣4WS and distal to the BsiWI near the 3Ј end in ␤2WS with the identical coding sequence from synthetic oligonucleotides up to and including Leu-242 (␣4WS NT ) or Lys-233 (␤2WS NT ) followed by a stop codon. The correctness of products derived from PCR or from chemically synthesized oligonucleotides was confirmed by DNA sequencing.
Protein Expression in Xenopus Oocytes-Xenopus laevis were purchased from NASCO (Fort Atkinson, WI) and maintained in 12-h light/ 12-h dark cycle at 18°C. The Institutional Animal Care and Use Committee of Texas A & M University approved all animal procedures. Oocytes were prepared following standard procedures (42,43), using collagenase type I (Invitrogen) for enzymatic stripping of the follicular layer. cRNA was synthesized from linearized plasmid DNA using an SP6 mMessage mMachine kit (Ambion, Austin, TX). A 40-nl aliquot containing 20 ng of cRNA for each subunit being studied was injected into the cytoplasm of each oocyte. The oocytes then were incubated at 18°C for 3-4 days in 50% Leibovitz's L-15 medium (Invitrogen) in 10 mM HEPES, pH 7.5, containing 10 units/ml penicillin and 10 g/ml streptomycin.
The membrane-bound fraction of subunit protein was extracted from oocyte membranes by homogenizing 25-50 oocytes by hand in ice-cold buffer A (in mM: 50 sodium phosphate, 50 NaCl, 5 EDTA, 5 EGTA, 5 benzamidine, 15 iodoacetamide, pH 7.5), centrifuging this FIGURE 1. Designs of the primary sequences of extracellular domain ␣4 and ␤2 subunits. ␣4M1, ␣4M1 NT , ␤2M1, and ␤2M1 NT contained all N-terminal residues of the full-length subunits through M1. In contrast, ␣4WS, ␣4WS NT , ␤2WS, and ␤2WS NT contained no transmembrane domain and potentially could form water-soluble extracellular domain nAChRs, leading to the inclusion of "WS" in the subunit names. Superscript NT designates subunit designs with no epitope tag. Numbering of the amino acid residues in the extracellular domain is shown in parentheses. In comparison, the total length of ␣4 is 627 amino acids; total length of ␤2 is 502 amino acids. The epitope tags for mAb 142 on ␣4M1 and mAb 236 on ␤2M1 are labeled 142 t and 236 t , respectively. The signal peptides are designated SP. Assignments of residues to topological and functional regions of the primary sequences were based on ␣4 and ␤2 entries in the Swiss-Prot data base. mixture for 30 min at 14,500 ϫ g and 4°C, solubilizing the pellet in buffer B (buffer A with 2% w/v Triton X-100) during gentle agitation for 2 h at 4°C, and again centrifuging for 30 min at 14,500 ϫ g and 4°C (44). The membrane fraction was defined as the detergent extract. The membrane fraction was used for immunoblotting and ligand binding assays of subunits with transmembrane domains. For ␣4WS, ␣4WS NT , ␤2WS, and ␤2WS NT , the secreted fraction from injected oocytes was defined as the L-15 medium incubating injected oocytes; the cytoplasmic fraction from injected oocytes was defined as the supernatant above the pellet after homogenizing oocytes in buffer A and centrifuging.
Immunoblotting-Membrane fractions of ␣4M1, ␤2M1, ␣4WS, and ␤2WS were deglycosylated with peptide:N-glycosidase F (PNGase F, New England Biolabs), a glycosylasparaginase, according to the manufacturer's instructions. Membrane fractions were processed in parallel without PNGase F. After deglycosylation, the samples were denatured by heating at 95°C for 90 s in 1% SDS and 15 mM dithiothreitol. The samples were run by SDS-PAGE on a 12.5% acrylamide gel and transferred to Immun-Blot TM polyvinylidene difluoride membrane (Bio-Rad) using a GENIE electrophoretic transfer device (Idea Scientific). The membrane was blocked overnight at 4°C with 5% powdered milk in PBS buffer and 0.05% Tween 20. Primary antibody (mAb 142 for ␣4M1 and ␣4WS; mAb 236 for ␤2M1 and ␤2WS) at 3 nM in blocking solution was incubated with the membrane overnight at 4°C, followed by washing for several hours with PBS and 0.05% Tween. Goat anti-rat secondary antibody conjugated with peroxidase (Pierce) was diluted 1:5000 in blocking buffer, incubated with the membrane overnight at 4°C, and then washed several hours with PBS and 0.05% Tween. Proteins on the membrane were detected on BioMax ML film (Eastman Kodak Co.) with SuperSignal West Dura Extended Duration Substrate (Pierce) according to the manufacturer's instructions.
Ligand Binding Assays-Immulon 4 HBX plastic microwells (Thermo Labsystems) for solid phase binding assays were coated with mAb 142 or mAb 236 for subunits with epitope tags (39,40); mAb 295, which binds the extracellular domain of ␤2 in its native conformation (45,46); or mAb 299 (Sigma), which binds the extracellular domain of ␣4 in its native or denatured conformation (45,47). The wells were blocked with 3% bovine serum albumin, fraction V (EM Science), in PBS (in mM: 137 NaCl, 2.7 KCl, 1.4 KH 2 PO 4 , 10 Na 2 HPO 4 ). For membrane-bound fractions, a volume of detergent extract containing the equivalent of 0.1 to 0.5 injected oocytes was added to each mAb-coated microwell. The microwells were incubated overnight at 4°C to allow binding of epitope-tagged subunits and then washed with ice-cold buffer B. To measure [ 3 H]epibatidine affinity, the microwells were loaded with 200 l of (Ϯ)-[ 3 H]epibatidine (PerkinElmer Life Sciences; specific activity ranged from 30 to 56 Ci/mmol) in ice-cold buffer B. To measure competitive inhibition by cytisine, (Ϫ)-nicotine, and acetylcholine (Sigma), the microwells were loaded with either 1 or 3 nM [ 3 H]epibatidine and varying concentrations of inhibitor in a total volume of 200 l of ice-cold buffer B. These two fixed concentrations of [ 3 H]epibatidine were chosen for the competitive inhibition assays, so depletion of free [ 3 H]epibatidine was negligible. For measuring [ 3 H]epibatidine affinity and for measuring competitive inhibition, the microwells were incubated overnight at 4°C to allow ligand binding. The microwells then were quickly washed with ice-cold buffer B and were stripped with a solution of 5% SDS and 25 mM dithiothreitol in buffer B. The amount of [ 3 H]epibatidine was measured in a liquid scintillation counter. Each data point was the average of duplicate or triplicate measurements. Microwells measuring [ 3 H]epibatidine binding from secreted fractions and cytoplasmic fractions of ␣4WS/␤2WS or ␣4WS NT / ␤2WS NT contained 5-10 oocyte equivalents during incubation with those fractions. Processing of these microwells was similar to processing of microwells measuring membrane-bound fractions, except buffer A replaced buffer B. Nonspecific binding was measured from uninjected oocytes and was less than 5% of specific binding at 30 nM total [ 3 H]epibatidine and was less than 1% of specific binding at concentrations below 1 nM total [ 3 H]epibatidine.

Models of [ 3 H]Epibatidine
Binding-Dissociation constants in the picomolar range have been reported for [ 3 H]epibatidine binding to ␣4␤2 nAChRs (48 -51). This high affinity raised the possibility of ligand depletion during binding assays. Ligand depletion means the free concentration of [ 3 H]epibatidine differs significantly from the total concentration (52). To estimate dissociation constants for [ 3 H]epibatidine, binding data were described with a one-site model and with cooperative and independent two-site models (Fig. 2). To allow for ligand depletion, the equations that described binding in the models were functions of the total concentration of [ 3 H]epibatidine and did not assume free and total concentrations of [ 3 H]epibatidine to be equal. Binding of [ 3 H]epibatidine for the one-site model was based on the mass action Equation 1 derived from the one-site binding mechanism in Fig. 2A, where R is the concentration of free ligand-binding sites (unbound receptor); E is the concentration of free where R t and E t are the total concentrations of receptor-binding sites and of [ 3 H]epibatidine, respectively. Equation 2 is derived from Equa- where RE is the total concentration of receptor with one bound [ In Equation 5, R t co is the total concentration of receptors (each receptor with two binding sites).
Equations 6 and 7 describe the concentration of [ 3 H]epibatidinebinding species for the two-independent sites model with where Z is the concentration of free [ 3 H]epibatidine and is the positive root of Equation 8, Inhibition of [ 3 H]epibatidine binding by cytisine, (Ϫ)-nicotine, and acetylcholine was described with a one-site competitive inhibition model (Fig. 2D). Total concentrations of [ 3 H]epibatidine for inhibition experiments were 1 and 3 nM. These concentrations were large enough that depletion of [ 3 H]epibatidine and inhibitors could be neglected. Equation 9 describes the concentration of the [ 3 H]epibatidine-binding species for the one-site competitive inhibition model, Based on the theory of ligand binding with ligand depletion (54,55), Equations 3-9 were derived with the program Maple (Maplesoft, Waterloo, Ontario, Canada) from the mass action equations for the two-site models and one-site competitive inhibition model, combined with the conservation equations for R t , R t , E t , and I t . Estimation of Dissociation Constants-K d for the one-site binding model was estimated with two methods. First, K d was calculated for each data set from the half-maximum concentration, K 0.5 , and total concentration of binding sites, R t , as determined from least squares fitting of Equation 10 to the data set and from Equation 11, In Equation 10, n is the Hill coefficient. The best estimate of K d with this method, which is based on the relationship between K d and K 0.5 in the one-site model (52,56), was the average of the K d values from all data sets. Second, K d was estimated with least squares fitting of Equation 2 simultaneously to all data sets of [ 3 H]epibatidine binding. The independent variable was E t . The value of K d was constrained to be identical for all data sets. Because the total concentration of receptor varied from data set to data set, a unique R t parameter value for each data set also was included in the simultaneous fitting to all data sets.
Values for K d1 co and K d2 co were estimated by least squares fitting of the equation RE ϩ 2 ϫ RE 2 (total concentration of bound [ 3 H]epibatidine) simultaneously to all data sets of [ 3 H]epibatidine binding, where RE and RE 2 are defined by Equations 3 and 4. The values of K d1 co and K d2 co were constrained to be identical for all data sets and were defined as microscopic or intrinsic dissociation constants and not statistical dissociation constants based on two equivalent binding sites per receptor (57). A unique parameter R t for each data set also was included in the simultaneous fitting, where R t ϭ RE ϩ RE 2 .
Values for K d1 in and K d2 in were estimated by least squares fitting of the equation from Equations 6 and 7 simultaneously to all relevant data sets. The values of K d1 in , K d2 in , and the ratio R t /R t (2) were constrained to be identical for all data sets. A unique R t parameter for each data set was included in the simultaneous fitting where R t ϭ R t (1) /R t (2) . K i , the dissociation constant for an inhibitor, was estimated with two methods based on the one-site competitive inhibition model (Fig. 2D). First, K i was calculated from the half-maximum concentration IC 50 as determined from fitting to Equation 12 and from the Cheng-Prusoff equation (58), Equation 13, where n is the Hill coefficient; I total is the total concentration of inhibitor; K d is the dissociation constant for [ 3 H]epibatidine, and C is a nonnegative constant to account for the base line. The best estimate of K i for this method was the average of the K i values across all data sets. As an extension of this method, a two-independent sites model of inhibition also was tested with Equation 14, Second, K i was estimated with least squares fitting of RE ϩ C, where RE is defined by Equation 9.
Equations 10 and 12 were fit to individual data sets with Kaleida-Graph (Synergy Software) to determine K 0.5 and IC 50 values. Simultaneous least squares fitting of Equations 2-4, 6, 7, and 9 to multiple data sets was done with the Solver tool in Microsoft Excel (59) (60). Quality of fit from different models was compared with an F test (61). Uncertainties are presented as standard error of the mean.
Sucrose Gradient Sedimentation-Detergent extracts of membranes from 3 to 5 oocytes injected with ␣4 and ␤2 subunits or with ␣4M1 and ␤2M1 subunits were processed as described for ligand binding assays. Membrane vesicles from TE671, a human rhabdomyosarcoma cell line (62) that contains muscle nAChRs, were solubilized in buffer B. A 200-l aliquot of each detergent extract was layered on a 5-ml sucrose gradient (5-20% (w/v)) in a 0.5% Triton solution (in mM: 100 NaCl, 10 sodium phosphate, 5 EGTA, 5 EDTA, 1 NaN 3 , pH 7.5). The gradients were centrifuged for 75 min at 70,000 rpm (340,000 ϫ g) and 4°C in a Beckman NVT90 rotor. Each gradient was collected in aliquots of 11 drops (130 l) into 40 microwells coated with mAb 236 for ␣4M1/ ␤2M1 nAChRs, mAb 295 for ␣4␤2 nAChRs, or mAb 210 (63) for nAChRs from TE671. After incubation at 4°C for 24 h that allowed for binding of protein to antibodies, the microwells were washed with buffer B and filled with 100 l of 2 nM [ 3 H]epibatidine for ␣4M1/␤2M1 and ␣4␤2 nAChRs or 4 nM 125 I-␣Bgt for muscle nAChRs in buffer B for 24 h at 4°C. Microwells were washed with buffer B. Bound [ 3 H]epibatidine was stripped from the microwells with a solution 5% SDS and 25 mM dithiothreitol in buffer B and measured in a liquid scintillation counter. Bound 125 I-␣Bgt was measured in a ␥-counter.
Comparison of nAChR Yields-The yield of [ 3 H]epibatidine-binding sites for each subunit combination was defined as the amount of bound [ 3 H]epibatidine with a total [ 3 H]epibatidine concentration of 1-3 nM. The absence of ligand depletion and absence of saturation of mAbbinding sites were confirmed by determining that the amount of bound [ 3 H]epibatidine was directly proportional to the amount of membrane fraction. Yield was calculated in two ways as follows: 1) the amount of [ 3 H]epibatidine-binding sites per oocyte and 2) the amount of [ 3 H]epibatidine-binding sites per oocyte relative to the yield of a standard subunit combination that was expressed and measured in parallel. The standards were ␣4␤2 or ␣4M1/␤2M1 nAChRs. Relative yield helped control for variations in protein expression that depended on time of year or the particular frog from which oocytes were harvested.

RESULTS
Protein Expression in Xenopus Oocytes-The ␣4M1, ␤2M1, ␣4WS, and ␤2WS subunits, when the corresponding cRNAs were injected as ␣4M1 and ␤2M1 or as ␣4WS and ␤2WS pairs, were glycosylated in oocytes (Fig. 3). The relatively narrow band of each glycosylated subunit suggested homogeneity in the attached carbohydrate structure. The molecular mass of each protein after deglycosylation with PNGase F, which cleaves N-linked glycoproteins, corresponded closely to the calculated value based on amino acid composition. The extracellular domain of ␣4 has three potential N-linked glycosylation sites (Asn-57, Asn-108, and Asn-174), based on an NX(S/T) motif; ␤2 has two potential sites (Asn-51 and Asn-168).
Binding Properties of [ 3 H]Epibatidine with ␣4M1/␤2M1 nAChRs-To begin analyzing the binding properties with [ 3 H]epibatidine, the affinities of ␣4␤2 nAChRs and ␣4M1/␤2M1 nAChRs were estimated according to the one-site model ( Fig. 2A). K d was estimated in two ways as follows: 1) from K 0.5 and R t values of the individual binding data sets according to Equation 11, and 2) by fitting Equation 2 simultaneously to all data sets (TABLE ONE). K 0.5 values, determined with Equation 10, ranged from 23 to 200 pM for ␣4␤2 nAChRs and from 20 to 100 pM for ␣4M1/␤2M1 nAChRs. The total concentration of [ 3 H]epibatidinebinding sites ranged from 23 to 400 pM for ␣4␤2 nAChRs and 34 to 160 pM for ␣4M1/␤2M1 nAChRs. The average K d values calculated from K 0.5 were 13 Ϯ 3 pM (n ϭ 8) for ␣4␤2 nAChRs and 8 Ϯ 2 pM (n ϭ 6) for ␣4M1/␤2M1 nAChRs. One data set for ␣4M1/␤2M1 nAChRs was not included in the estimate because its calculated K d value based on K 0.5 and R t for that data set was negative. The K d values from simultaneous fitting of all data sets to the one-site model were similar to the estimates based on K 0.5 values: K d ϭ 16 pM (95% CI, 7-29 pM, n ϭ 8) for ␣4␤2 nAChRs and K d ϭ 10 pM (95% CI, 5-17 pM, n ϭ 7) for ␣4M1/␤2M1 nAChRs. To determine whether a two-sites model was better than the one-site model, models of two cooperative sites (Fig. 2B) and two independent sites (Fig. 2C) were fit to the data. For ␣4␤2 nAChRs, the best values for the dissociation constants for the two-cooperative sites model were K d1 co ϭ 23 pM and K d2 co ϭ 14 pM. This cooperative model, however, did not fit the data better than the one-site model (p ϭ 0.54). For ␣4M1/␤2M1 nAChRs, the best values were K d1 co ϭ 4.8 pM and K d2 co ϭ 15 pM; however, this cooperative model did not fit the data significantly better than the one-site model (p ϭ 0.13). These results showed the two-cooperative sites model was not better than the one-site model for either ␣4␤2 or ␣4M1/␤2M1 nAChRs.
The two-independent sites model for ␣4␤2 nAChRs contained a large fraction of a high affinity site (picomolar for K d1 in ) and a small fraction of a low affinity site (nanomolar for K d2 in ) (TABLE ONE). The best values for the dissociation constants were K d1 in ϭ 13 pM (population fraction, 0.84; 95% CI ϭ 3.4 -23 pM at fixed population ratios) and K d2 in ϭ 12 nM (population fraction, 0.16; 95% CI ϭ 11 pM, no upper limit at fixed population ratios). The value of K d1 in was similar to the value of K d in the one-site model (16 pM). The two-independent sites model fit the data better than the one-site model (p ϭ 0.002). When the analysis was restricted to the three data sets with 30 nM maximum total [ 3 H]epibatidine, K d1 in ϭ 4.9 pM (population fraction, 0.76) and K d2 in ϭ 0.59 nM (population fraction, 0.24). The two-independent sites model fit these three data sets better than the one-site model (p ϭ 0.0001). A threeindependent sites model did not fit the data better than the two-independent sites model (p ϭ 0.9997).
The two-independent sites model for ␣4M1/␤2M1 nAChRs contained one site with a large population fraction and picomolar affinity and a second site with a small population fraction with subnanomolar affinity: K d1 in ϭ 7.4 pM (population fraction, 0.91) and K d2 in ϭ 0.35 nM (population fraction, 0.09) (TABLE ONE). The value of K d1 in was similar to the value of K d in the one-site model (10 pM). The two-independent sites model, however, did not fit the data significantly better than the one-site model (p ϭ 0.09). When the analysis was restricted to the three data sets with a maximum of 30 nM total [ 3 H]epibatidine, then K d1 in ϭ 8.0 pM (population fraction, 0.95) and K d2 in ϭ 3.0 nM (population fraction, 0.05). The two-independent sites model, however, did not fit these three data sets significantly better than the one-site model (p ϭ 0.09).
In summary, the two-independent sites model provided the best fit for ␣4␤2 nAChRs, and the one-site model provided the best fit for ␣4M1/␤2M1 nAChRs (Fig. 4). The high affinity site contributed a large majority of the [ 3 H]epibatidine-binding sites for ␣4␤2 nAChRs. The value of K d1 in (13 pM) for the high affinity site on ␣4␤2 nAChRs was similar to the value of K d (10 pM) for the single site on ␣4M1/␤2M1 nAChRs.
To determine whether the presence of the epitope tag at the C terminus of the ␣4M1 and ␤2M1 affected the binding properties of ␣4M1/ ␤2M1 nAChRs with [ 3 H]epibatidine, binding was measured with ␣4M1 NT and ␤2M1 NT , which did not have epitope tags. K d values for ␣4M1 NT /␤2M1 NT nAChRs estimated from either K 0.5 and R t or from the single site model were similar to K d values for ␣4M1/␤2M1 nAChRs and for ␣4␤2 nAChRs (TABLE ONE). The two-cooperative sites model did not fit the data significantly better than the one-site model (p ϭ 0.07). In contrast, the two-independent sites model fit significantly better than the one-site model (p ϭ 0.0005). K d1 in for ␣4M1 NT /␤2M1 NT nAChRs was comparable with K d1 in for ␣4M1/␤2M1 nAChRs and characterized 85% of the binding sites. K d2 in characterized 15% of the binding sites and showed nanomolar affinity. These results showed the epitope tag was not necessary for forming the major high affinity site, although the tag might modify the heterogeneity of the binding sites. The major binding sites from ␣4␤2 nAChRs, ␣4M1/␤2M1 nAChRs, and ␣4M1 NT / ␤2M1 NT nAChRs showed similar affinities for [ 3 H]epibatidine (TABLE  ONE).
To determine whether binding properties for ␣4M1/␤2M1 nAChRs depended on the presence of mAb 142 in the assay, ␣4M1/␤2M1 nAChRs were immunoisolated on microwells with three other antibodies as follows: mAb 295, mAb 236, and mAb 299 (Fig. 5). Values of K d estimated from either K 0.5 and R t or from fitting the one-site model to binding data with a maximum of 30 nM total [ 3 H]epibatidine ranged from 7 to 11 pM (TABLE TWO) and were similar to the value of K d with mAb 142 (10 pM). The two-independent sites model did not fit these data better than the one-site model (p Ͼ 0.28 for all mAbs). No site, therefore, with nanomolar affinity was identified with any of the three  Fig. 2  in values for ␣4␤2 nAChRs, the K d value for ␣4M1/␤2M1 nAChRs, and total receptor concentration for each data set (R t ) were determined by fitting the models to all data sets simultaneously. Each line was calculated with the total receptor concentration that best fit the corresponding data set. Solid lines are fits to data displayed with filled and half-filled squares and upright and leftward triangles. Dotted lines are fits to data displayed with circles, diamonds, and inverted and rightward triangles. In the ␣4M1/␤2M1 panel, only six lines can be seen, even though seven data sets appear. The solid line fitting data displayed as leftward triangles overlaps and obscures a dashed line fitting data displayed as rightward triangles. The maximum concentration of total [ 3 H]epibatidine ranged from 1 to 30 nM. DECEMBER 2, 2005 • VOLUME 280 • NUMBER 48 mAbs. These results with different mAbs showed that binding affinity of ␣4M1/␤2M1 nAChRs for [ 3 H]epibatidine did not depend on the antibody used for immunoisolation.

Extracellular Domain ␣4␤2 Nicotinic Acetylcholine Receptors
Binding Properties of Cytisine, Nicotine, and ACh with ␣4M1/␤2M1 nAChRs-Binding properties of ␣4M1/␤2M1 nAChRs and ␣4␤2 nAChRs were further compared with cytisine, nicotine, and ACh. The dissociation constants for these ligands were estimated with the following two methods based on their competitive inhibition of binding by [ 3 H]epibatidine: 1) from IC 50 values based on fitting the data with Equation 12 and calculations with the Cheng-Prusoff equation (Equation 13), and 2) from a one-site competitive inhibition model with or without inhibitor depletion (Fig. 2D). Even though the two-independent sites model provided the best fit to [ 3 H]epibatidine binding data with ␣4␤2 nAChRs, the one-site competitive inhibition model was used with competitive inhibition assays. This simplification was based on the assumption that the large population fraction of the high affinity site dominated the competitive inhibition data.
The values of dissociation constants of cytisine, nicotine, and ACh with ␣4M1/␤2M1 nAChRs were similar to the values with ␣4␤2 nAChRs (TABLE THREE), whether the estimates were based on IC 50 values and the Cheng-Prusoff equation (Fig. 6) or the one-site competitive inhibition model (Fig. 7). A two-independent sites model of inhibition (Equation 14) did not fit the data significantly better for either ␣4␤2 nAChRs (p Ͼ 0.15) or ␣4M1/␤2M1 nAChRs (p Ͼ 0.36) for any of these ligands. Estimates of dissociation constants did not depend signif-icantly on whether the total [ 3 H]epibatidine concentration was 1 or 3 nM.
All three inhibitors displaced nearly all [ 3 H]epibatidine bound to ␣4M1/ ␤2M1 nAChRs at the highest concentrations of inhibitor (Fig. 6). The base line from Equation 12 as a fraction of binding without inhibitor and with 1 nM [ 3 H]epibatidine was 0.0038 (CI, 0 -0.055) for cytisine, 0.017 (CI,  Fig. 2A and Equation 2) to a specific data set. For these calculations, K d and total receptor concentrations for each data set (R t ) were determined by fitting the model to all data sets simultaneously for a given mAb. Each line was calculated with the total receptor concentration that best fit the corresponding data set. The dissociation constant did not depend significantly on which mAb immunoisolated ␣4M1/␤2M1 nAChRs (TABLE  TWO).  To determine whether the antibody used to immunoisolate ␣4M1/ ␤2M1 nAChRs affected the inhibition assays, the K i value for nicotine was estimated with mAb 295, mAb 236, and mAb 299. The values of K i from the one-site inhibition model did not significantly depend on the mAb (TABLE TWO). The base line, however, did depend on the mAb, with the base line for mAb 295 being significantly different from zero (Fig. 9). The average base lines as fractions of the binding without inhibitor were 0.34 (95% CI, 0.15-0.46) for mAb 295, 0.00 (95% CI, 0.00 -0.15) for mAb 236, and 0.00 (95% CI, 0.00 -0.14) for mAb 299. For both ␣4␤2 nAChRs and ␣4M1/␤2M1 nAChRs, immunoisolation by mAb 295 was associated with incomplete inhibition of [ 3 H]epibatidine binding. In contrast, immunoisolation by mAb 142, mAb 236, and mAb 299 was associated with complete inhibition of [ 3 H]epibatidine binding.
Oligomerization of ␣4M1 and ␤2M1 Subunits-[ 3 H]Epibatidinebinding sites arose from a heteromeric combination of ␣4M1 and ␤2M1 subunits because neither ␣4M1 nor ␤2M1 subunits alone produced detectable [ 3 H]epibatidine binding. To estimate the number of subunits in ␣4M1/␤2M1 nAChRs, the molecular mass of ␣4M1/␤2M1 nAChRs was estimated from sedimentation on sucrose gradients (Fig. 10A). The size standards were pentameric ␣4␤2 nAChRs and pentameric (␣1) 2 ␤␥␦ nAChRs and ␣1 subunit monomers from TE671, a human rhabdomyosarcoma cell line. The peak from ␣4␤2 nAChRs ran slightly faster than the peak from TE671 nAChRs, as was predicted from their molecular masses of 297 kDa for ␣4␤2 nAChRs and 268 kDa for human muscle nAChRs. The peak of the monomeric ␣1 subunit, although broad, was near the top of the gradient, as was predicted from its molecular mass of 52 kDa. The distribution of [ 3 H]epibatidine binding by ␣4M1/␤2M1 nAChRs was narrow and formed a single peak, suggesting that one multimeric species contributed a significant fraction of the binding sites. In contrast, binding by a mixture of oligomeric states with significant fractions of dimers, trimers, tetramers, pentamers, and higher order aggregates would be expected to form a broad distribution or several different peaks. The molecular mass of the ␣4M1/␤2M1 nAChRs as estimated from the size standards was 170 kDa (Fig. 10A,  inset). The calculated molecular mass of a pentamer with a stoichiometry of (␣4M1) 2 (␤2M1) 3 was 154 kDa without glycosylation. This estimate for molecular mass, therefore, is consistent with a pentamer predominantly but not necessarily exclusively forming the high affinity [ 3 H]epibatidine-binding sites.
Time-dependent chemical cross-linking of ␣4M1 and ␤2M1 subunits expressed together provided additional evidence for oligomerization into pentamers (Fig. 10B). The products of cross-linking were denatured and reduced and analyzed by immunoblotting with mAb 142. A ladder of ␣4M1-containing protein bands of increasing apparent molecular masses formed as the cross-linking reaction proceeded. This time-dependent ladder was consistent with formation initially of dimers and trimers, then tetramers, and ultimately pentamers by cross-linking between subunits of ␣4M1/␤2M1 pentamers. The disappearance of lower molecular mass   DECEMBER 2, 2005 • VOLUME 280 • NUMBER 48 bands with increasing time suggested that most ␣4M1 subunits in oligomers were present in pentamers in the cross-linking reaction.

Extracellular Domain ␣4␤2 Nicotinic Acetylcholine Receptors
Yield of ␣4M1/␤2M1 nAChRs-To evaluate the efficiency of expressing extracellular domain nAChRs, yields of nAChRs from derivatives of ␣4 and ␤2 subunits were measured in absolute amounts of [ 3 H]epibatidine-binding sites and in amounts relative to binding sites from ␣4␤2 nAChRs or ␣4M1/␤2M1 nAChRs assayed in parallel (Fig. 11). Evaluating relative amounts helped control for seasonal variation and frog-tofrog variation of oocyte expression. The averages of absolute amounts varied 10-fold from 46 (␣4 and ␤2M1) to 440 fmol/oocyte (␣4M1 NT and ␤2M1 NT ). The relative amounts, however, varied about 2-fold across the different subunit combinations, ranging from 0.32 (␣4M1 and ␤2M1 NT ) to 0.63 (␣4M1 and ␤2) relative to ␣4␤2 nAChRs. Both ␣4WS and ␤2WS proteins were expressed on oocytes and were glycosylated (Fig. 3). In contrast to ␣4M1 and ␤2M1 subunits, however, ␣4WS and ␤2WS did not produce a detectable amount of [ 3 H]epibati- was zero (95% CI, 0 -0.14) with mAb 299. In contrast, the base line was significantly different from zero with mAb 295 (Fig. 6 and Fig. 7). Total [ 3 H]epibatidine concentration was 1 nM.  ) was fit to the same data sets as were analyzed for Fig. 6 but without normalization of the data sets. K i and total receptor concentrations for each data set (R t ) were determined by fitting the model to all data sets simultaneously for a given inhibitor. Each line was calculated with the total receptor concentration that best fit the corresponding data set. Total [ 3 H]epibatidine concentration was 1 nM.
dine-binding sites in the growth medium (secreted fraction), cytosol (cytoplasmic fraction), or Triton-extracted membranes (membrane fraction). The lower limit of detection of [ 3 H]epibatidine-binding sites in these assays with mAb 142-coated or mAb 236-coated microwells was estimated to be about 0.08 fmol/oocyte. Hence, the ratio of the yield of ␣4WS/␤2WS nAChRs to yield of ␣4M1/␤2M1 nAChRs was less than 0.001. To determine whether the C-terminal epitope tags interfered with production of ␣4WS/␤2WS nAChRs, ␣4WS NT and ␤2WS NT subunits, which did not contain epitope tags (Fig. 1), were expressed together in oocytes and tested for [ 3 H]epibatidine binding. With mAb 295-coated microwells, no [ 3 H]epibatidine-binding sites were detected in the secreted, cytoplasmic, or membrane fractions. These results showed that factors other than the epitope tags were responsible for the failure to detect ␣4WS/␤2WS nAChRs.

DISCUSSION
The nearly identical ligand affinities of ␣4M1/␤2M1 nAChRs, the predominant species for ␣4␤2 nAChRs, and the assembly of ␣4M1 and ␤2M1 subunits into multimers that likely are pentamers suggest that extracellular domains with the M1 domains of ␣4 and ␤2 subunits form high fidelity structural models of ␣4␤2 nAChRs. The values of dissociation constants for binding between ␣4␤2 nAChRs and [ 3 H]epibatidine, cytisine, nicotine, and acetylcholine are similar to values reported from other laboratories (48 -51, 65-67). A two-site model of binding of [ 3 H]epibatidine that previously was applied to ␣4␤2 nAChRs expressed in Xenopus oocytes and immunoisolated with mAb 299 found dissociation constants of 14 pM (population fraction, 0.73) and 1.2 nM (population fraction, 0.27) (49). Based on the relatively low precision of the estimated value of K d2 in for ␣4␤2 nAChRs, the difference between 1.2 and 12 nM for these two estimates of the dissociation constant of the low affinity [ 3 H]epibatidine site was not significant.
With respect to design requirements for extracellular domain ␣4␤2 nAChRs, the extracellular domains with the first transmembrane domains of ␣4 and ␤2 subunits were sufficient for assembly into extracellular domain ␣4M1/␤2M1 nAChRs that were stable after solubilization in detergent. The cytoplasmic loop was not required for assembly or for stability of these extracellular domain nAChRs after solubilization. M1, however, played an important role for expressing these extracellular domain nAChRs. These results, with the properties of extracellular domain ␣7 nAChRs (33,34) and evidence of oligomerization of extracellular domains from ␣1 and ␦ subunits (31,32), from extracellular domains of muscle-type subunits (35), from chimeric ␣7/AChBP subunits (26), from glycine receptor subunits (68,69), and from GABA A receptor subunits (70), suggest that producing extracellular domain receptors might be possible for many subunits throughout the nicotinoid family. Expression of extracellular domain nicotinoid receptors with high structural fidelity, however, might be more feasible with M1 included in the subunit design than without M1.
Conclusions about the extent of structural similarity between ␣4M1/ ␤2M1 and ␣4␤2 nAChRs have limitations. First, the data for [ 3 H]epibatidine binding to ␣4␤2 nAChRs were fit better by the two-independent sites model than by the one-site model, implying a heterogeneous population of binding sites. In contrast, the data for [ 3 H]epibatidine binding to ␣4M1/␤2M1 nAChRs were fit better by the one-site model, suggesting that ␣4M1/␤2M1 nAChRs model only a subset of different ␣4␤2 nAChR structures. One possible cause of a heterogeneous popu- The inset shows a size calibration curve with positions of peaks from the standard nAChRs (pentameric ␣4␤2 nAChRs, pentameric fetal muscle nAChRs, and monomeric ␣1 from bottom to top of gradient) as circles, and the position of the peak from ␣4M1/␤2M1 nAChRs as a cross. The calibration curve was based on the calculated molecular masses of the three standard proteins without post-translational modification and assumed a subunit stoichiometry for ␣4␤2 nAChRs of (␣4) 2 (␤2) 3 . The estimated molecular mass at the peak from ␣4M1/␤2M1 nAChRs was 170 kDa. By comparison, the predicted molecular mass of pentameric (␣4M1) 2 (␤2M1) 3 nAChRs without post-translational modification was 154 kDa. This estimated molecular mass suggested that ␣4M1/␤2M1 nAChRs predominantly were pentamers. B, detergent extract of ␣M1 and ␤2M1 subunits expressed together in Xenopus oocytes was chemically cross-linked for 10, 30, 60, and 120 min, denatured, and visualized by immunoblotting with mAb 142. Bands containing ␣4M1 appeared as a ladder of increasing apparent molecular masses; higher molecular masses formed as the cross-linking reaction proceeded. As labeled in the figure, the apparent molecular masses of these bands and the relative timing of their formation suggested the sequential cross-linking of ␣4M1-containing oligomers to form dimers (about 80 kDa), trimers (about 120 kDa), tetramers (about 160 kDa), and pentamers (about 190 kDa). These results are consistent with the expected products from cross-linking of pentameric ␣4M1/␤2M1 nAChRs. DECEMBER 2, 2005 • VOLUME 280 • NUMBER 48 lation of binding sites in ␣4␤2 nAChRs could be the presence of two or more relative subunit stoichiometries between ␣4 and ␤2 subunits. The ␣4M1:␤2M1 ratio and ␣4:␤2 ratio are likely 2:3, based on evidence from chicken ␣4␤2 nAChRs expressed in oocytes with a 1:1 ratio of injected ␣4 and ␤2 cRNAs (71) or with a 1:1 ratio of nuclear-injected cDNAs (72). In addition to the 2:3 ratio, however, an ␣4:␤2 ratio of 3:2 in ␣4␤2 nAChRs has been found in human embryonic kidney tsA201 cells (73). Other ␣4:␤2 ratios besides 2:3 also have been implied by pharmacologic properties of ␣4␤2 nAChR expressed from different ratios of injected ␣4 and ␤2 subunits (74,75).

Extracellular Domain ␣4␤2 Nicotinic Acetylcholine Receptors
Second, additional evidence suggesting structural heterogeneity in ␣4␤2 nAChRs as well as in ␣4M1/␤2M1 nAChRs was incomplete inhibition of [ 3 H]epibatidine binding by cytisine, nicotine, and acetylcholine that was associated with immunoisolation by mAb 295. In contrast, complete inhibition of [ 3 H]epibatidine binding was associated with immunoisolation by mAb 142, mAb 236, and mAb 299. These differences in inhibition behavior might identify minor populations of nAChRs captured by mAb 295 and possessing binding properties distinguishable from the dominant population. These binding properties might arise from unusual subunit stoichiometries or post-translational modifications or might be associated with immature subunit assemblies. A protein population with structural homogeneity facilitates structural analysis. Therefore, determining the range of relative subunit stoichiometries of extracellular domain ␣4␤2 nAChRs and developing methods for purifying populations with homogeneous subunit composition and structure will be important steps in the future for developing these nAChRs as structural models.
Third, similarity of ligand dissociation constants does not require global structural fidelity of the ␣4M1/␤2M1 nAChRs with the extracellular domain of ␣4␤2 nAChRs. In other words, structural fidelity at the ligand-binding site might coexist with structural differences outside that site. Similar ligand binding properties likely reflect similar global structures in nAChRs, however, because the ligand-binding site likely is formed by widely dispersed, noncontiguous sections of the primary sequence, based on the structure of AChBP (9,76).
Similar to the effect of M1 on expression of extracellular ␣7 nAChRs (34), the presence of M1 in both ␣4M1 and ␤2M1 appears to greatly increase the efficiency of forming extracellular ␣4␤2 nAChRs in oocytes. This conclusion is based on detecting [ 3 H]epibatidine-binding sites formed with ␣4M1 and ␤2M1 but not with ␣4WS and ␤2WS. What hindered the expression of ␣4WS/␤2WS nAChRs? The failure of ␣4WS and ␤2WS to form extracellular domain nAChRs was not caused by failure to synthesize these proteins or by excessive instability of these proteins. They were synthesized and were stable enough to be detected by immunoblotting at levels in detergent extract similar to the levels ␣4M1 and ␤2M1. The C-terminal epitope tags were not the sole block to forming extracellular domain nAChRs because removing these epitope tags from ␣4WS and ␤2WS did not lead to expression of [ 3 H]epibatidine-binding sites. A possible explanation is that the truncation points in the primary sequences of ␣4 and ␤2, chosen because those points juxtaposed with M1, might be unsuitable for forming extracellular domain ␣4␤2 nAChRs. In that case, shorter designs of ␣4WS and ␤2WS might be better suited for expressing water-soluble extracellular domain ␣4␤2 nAChRs. Increasing the hydrophilic character of the Cys loop in the extracellular domain is another potentially favorable change in the designs of ␣4WS and ␤2WS, based on the increased water solubility of a chimeric ␣7 extracellular domain with a more hydrophilic Cys loop sequence taken from AChBP (26).
Why might M1 be necessary for efficiently forming extracellular domain nAChRs? Previous speculation has centered on a possible role of M1 as a tether to membrane surfaces, which might reduce the degrees of spatial freedom within and between subunits and favor correct folding and assembly (31,32). Such tethering might also favor interaction with membrane-bound host components of the endoplasmic reticulum or Golgi that contribute to formation of nAChRs. The presence of a sequence motif in the M1 of ␣1 that governs surface trafficking (77) supports an alternative possibility that M1 is not acting simply as a membrane tether. Instead, properties specific to M1 and not found in other transmembrane domains might be important for formation of extracellular domain nAChRs.
With evidence for the importance of M1 to extracellular domain nAChRs, how can water-soluble extracellular domain nAChRs be made efficiently? First, they might be produced by in vitro proteolysis that removes transmembrane domains from nAChRs. This strategy would require engineering of specific proteolysis sites into subunit sequences and has been explored with ␣7 (34). Second, concatenated subunits without transmembrane domains might form nAChRs without a membrane surface. Formation of nAChRs from concatenated ␣4 and ␤2 units with transmembrane domains (78) supports the feasibility of this approach. Third, chimeric extracellular domain nAChR subunits that include sequences from AChBP might lead to water-soluble pentamers with high structural fidelity to the parent full-length nAChRs. Studies of chimeric, water-soluble extracellular domains from ␣7 and AChBP sequences (26) and chimeric ion channels containing a chimeric AChBP/5-hydroxytryptamine (serotonin) receptor subunit type 3A extracellular domain and the pore domain from the 5-hydroxytryptamine (serotonin) receptor subunit type 3A subunit (79) have begun to explore the feasibility of this approach. Fourth, increasing host factors that contribute to subunit folding and assembly also might improve the efficiency of producing water-soluble nAChRs. Properties of expression systems that promote, permit, hinder, or prevent forming of extracellular nAChRs need further investigation.
Even if water-soluble nAChRs do not prove to be feasible for structural studies, they also might be unnecessary in the future. For example, one advantage of extracellular domain nAChRs is their smaller molecular mass compared with full-length nAChRs. With or without M1, their smaller size will make them better candidates than full-length nAChRs for NMR spectroscopy as the upper size limits for protein NMR continue to increase through technological advances. Presently, these size limits are from 40 to 80 kDa for monomeric water-soluble proteins (80,81), up to 900 kDa for homo-oligomeric water-soluble proteins (82), and about 20 kDa for monomeric membrane proteins (83,84). Eventually, describing structural and dynamic properties of ligand binding, channel gating, and ion permeation will require high resolution structures of full-length nAChRs. Extracellular domain nAChRs are potentially important and accessible contributors to this developing story.