Assembly of human neuronal nicotinic receptor alpha5 subunits with alpha3, beta2, and beta4 subunits.

Nicotinic acetylcholine receptors formed from combinations of α3, β2, β4, and α5 subunits are found in chicken ciliary ganglion neurons and some human neuroblastoma cell lines. We studied the co-expression of various combinations of cloned human α3, β2, β4, and α5 subunits in Xenopus oocytes. Expression on the surface membrane was found only for combinations of α3β2, α3β4, α3β2α5, and α3β4α5 subunits but not for other combinations of one, two, or three of these subunits. α5 subunits assembled inside the oocyte with β2 but not with α3 subunits or other α5 subunits. α5 subunits coassembled very efficiently with α3β2 or α3β4 combinations. The presence of α5 subunits had very little effect on the binding affinities for epibatidine of receptors containing also α3 and β2 or α3 and β4 subunits. The presence of α5 subunits increased the rate of desensitization of both receptors containing also α3 and β2 or α3 and β4 subunits. In the case of receptors containing α3 and β4 subunits, the addition of α5 subunits had little effect on the responses to acetylcholine or nicotine. However, in the case of receptors containing α3 and β2 subunits, the addition of α5 subunits reduced the EC50 for acetylcholine from 28 to 0.5 μM and the EC50 for nicotine from 6.8 to 1.9 μM, while increasing the efficacy of nicotine from 50% on α3β2 receptors to 100% on α3β2α5 receptors. Both α3β2 and α3β2α5 receptors expressed in oocytes sedimented at the same 11 S value as native α3-containing receptors from the human neuroblastoma cell line SH-SY5Y. In the receptors from the neuroblastoma α3, β2, and α5 subunits were co-assembled, and 56% of the receptor subtypes containing α3 subunits also contained β2 subunits. The β2 subunit-containing receptors from SH-SY5Y cells exhibited the high affinity for epibatidine characteristic of receptors formed from α3 and β2 or α3, β2, and α5 subunits rather than the low affinity exhibited by receptors formed from α3 and β4 or α3, β4, and α5 subunits. Nicotine, like the structurally similar toxin epibatidine, also distinguishes by binding affinity two subtypes of receptors containing α3 subunits in SH-SY5Y cells. The affinities of α3β2 receptors expressed in oocytes were similar to the affinities of native α3 containing receptors from SH-SY5Y cells for acetylcholine, cytisine, and 1,1-dimethyl-4-phenylpiperazinium.

structurally similar toxin epibatidine, also distinguishes by binding affinity two subtypes of receptors containing ␣3 subunits in SH-SY5Y cells. The affinities of ␣3␤2 receptors expressed in oocytes were similar to the affinities of native ␣3 containing receptors from SH-SY5Y cells for acetylcholine, cytisine, and 1,1-dimethyl-4-phenylpiperazinium.
Nicotinic acetylcholine receptors (AChRs) 1 are members of a gene superfamily of homologous ligand-gated ion channels which include receptors for glycine, ␥-aminobutyric acid, and serotonin (1). There are three branches of the AChR gene family (2)(3)(4)(5). The best characterized are muscle and electric organ AChRs which consist of a pentameric array of homologous subunits oriented around a central ion channel like barrel staves. The order of these subunits around the channel is ␣1␥␣1␦␤1 in the fetal form and ␣1⑀␣1␦␤1 in the adult form (6). The two ligand binding sites in each AChR are thought to be formed at the interfaces between ␣1 and ␥, ␦, or ⑀ subunits (6). One group of neuronal AChRs which is capable of functioning as homomers is formed of ␣7, ␣8, or ␣9 subunits (although naturally occurring heteromers of ␣7 with ␣8 subunits have been described (7)). The remaining group of neuronal AChRs requires at least two kinds of subunits (2,8). It has been shown that ␣2, ␣3, and ␣4 subunits can form functional AChRs when expressed in pairwise combination with ␤2 or ␤4 subunits (9), suggesting that the ACh-binding sites are formed at specific interfaces between these ␣ and ␤ subunits. AChRs of the predominant brain subtype with high affinity for nicotine, when expressed in Xenopus oocytes from cloned subunits, have been shown to have a pentameric subunit composition with the stoichiometry (␣4) 2 (␤2) 3 (10,11). ␣3 AChRs of chicken ciliary ganglion have been shown to consist of 80% AChRs with the subunit composition ␣3␤4␣5 and 20% of AChRs with the subunit composition ␣3␤2␤4␣5 (8). ␣5 subunits have a cysteine pair homologous to ␣1 cysteines 192,193 which are located near the ACh-binding site of ␣1 subunits. This pair of cysteines accounts for their designation as ␣ subunits (12)(13)(14), but several putative ligand binding site amino acids are not conserved between ␣5 and other ␣ subunits (e.g. ␣5 lacks two critical tyrosines labeled by competitive antagonists (15,16)), ␣5 subunits are most closely related in sequence to ␤3 subunits (17), and ␣5 subunits do not form functional AChRs as homomers or in paired combination with ␤1, ␤2, or ␤4 subunits (12,13). Thus, like ␤1 subunits, ␣5 subunits may not be able to form ACh-binding sites by assembling with the appropriate inter-* This work was supported in part by a grant from the Pittsburgh Supercomputing Center through National Institutes of Health National Center for Research Resources cooperative agreement 1 p41 RR06009. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. face of other subunits, perhaps leading to the sort of subunit organization depicted in Fig. 1. Although recently it has been reported that co-expression of ␣5 with ␣4 and ␤2 subunits produced changes in conductance states and lower affinity for several agonists (18), studies of ␣5 heterologously expressed in combination with ␣3, ␤2, and ␤4 subunits have not been reported. Here we report such studies using human ␣3, ␤2, ␤4, and ␣5 subunits expressed in Xenopus oocytes.
The subunit compositions of native human ␣3 AChRs also need to be defined. The human peripheral neuroblastoma cell line SH-SY5Y resembles fetal sympathetic neurons in culture (19). Like chick ciliary ganglion neurons (8), SH-SY5Y cells express mRNAs for ␣3, ␣5, ␣7, ␤2, and ␤4 subunits (20,21). Ciliary ganglion AChRs assemble postsynaptic and perisynaptic AChRs from ␣3, ␤2, ␤4, and ␣5 subunits and perisynaptic AChRs from ␣7 subunits (8,22). Similarly, SH-SY5Y cells express postsynaptic type ␣3 AChRs which do not bind ␣-bungarotoxin and ␣7 type AChRs which do bind ␣-bungarotoxin (20,21). We use the ␣3 AChRs of SH-SY5Y cells as models of native human ganglionic ␣3 AChRs to compare with the properties of cloned human ␣3 AChR subtypes expressed in Xenopus oocytes. EXPERIMENTAL PROCEDURES cDNAs, mAbs, and Antisera-The cDNA sequences for human ␣3 (unpublished) and ␤2 (23) were submitted to EMBL (accession numbers X53559 and X53179, respectively). They were subcloned in expression vectors pcDNAI (Invitrogen) and pSP64poly(A) (Promega), respectively. The cDNA for human ␣5 was first described by Chini et al. (14) and kindly provided by Dr. Clementi (University of Milan). It was subcloned in the pSP64poly(A) vector. The cDNA for human ␤4 was cloned in this laboratory from a cDNA library from the neuroblastoma cell line SH-SY5Y. 2 It was then subcloned into the pcDNAI vector for in vitro expression. Tagging of human ␣3 (␣3 t ) with a "reporter epitope" at its C terminus was done by inserting three pairs of oligonucleotides which code for the peptide VSISPESDRPDLSTFGGVSISPESDRPDLST-FGSVSISPESDRPDLSTF containing 3 copies of the mAb236 epitope VSISPESDRPDLSTF (24). The insertion is between the restriction sites NsiI (in human ␣3 cDNA sequence, overlapping with its original stop codon) and XbaI (in vector pcDNAI sequence) sites of the cDNA clone H␣3/pcDNAI. Tagging of human ␣5 (␣5 t ) with a reporter epitope at its C terminus was done by inserting two pairs of oligonucleotides which code for the peptide SQVTGEVIFQTPLIKNPLQQVTGEVIFQTP-LIKNPLQ containing 2 copies of the mAb142 epitope QVTGEVIFQTP-LIKNP (24). The insertion is between the restriction sites AseI (in the human ␣5 cDNA sequence, 28 nucleotides upstream of its original stop codon) and SacI (in the vector pSP64poly(A) sequence) sites of the cDNA clone H␣5/pSP64poly(A). Monoclonal antibody mAb 210 was initially described as being directed at the main immunogenic region on the extracellular surface of ␣1 subunits (25) as was mAb35 (26). mAb35 binds to chick ␣3 AChRs (8). Similarly, we show here that mAb210 can cross-react with native human ␣3 and ␣5, but not ␤2 or ␤4 subunits. mAb290 to ␤2 subunits was initially described by Whiting and Lindstrom (27). It does not cross-react with ␣3, ␤4, and ␣5 subunits. mAb268 to ␣5 subunits was first reported by Whiting et al. (28) and further characterized by Conroy et al. (29). It binds to denatured but not to native ␣5 subunits. mAb142 and mAb236 to the ␣1 subunit were described initially by Tzartos et al. (30) and Criado et al. (31). Both mAb142 and mAb236 have been used effectively to detect reporter epitopes (24). Rabbit antiserum 3709 was raised against a synthetic peptide corresponding to a unique part of the human ␣3 subunit large cytoplasmic domain (348-NLNCFSRAESKGCKEGYPCGDGMCGYCH-HRRIK-380). It does not cross-react with human AChR subunits ␤2, ␤4, or ␣5 t on Western blots. Rabbit antiserum 3724 was raised against a synthetic peptide corresponding to a unique part of the human ␤2 subunit large cytoplasmic domain (387-GPGRSGEPCGCGLRE-401). It does not cross-react with human AChR subunits ␣3, ␤4, or ␣5 t on Western blots.
Purification and Immunoabsorption of AChRs from Oocytes and SH-SY5Y Cells-Oocytes were homogenized by repetitive pipetting in buffer A (50 mM Na 2 HPO 4 -NaH 2 PO 4 , pH 7.5, 50 mM NaCl, 5 mM EDTA, 5 mM EGTA, 5 mM benzamide, 15 mM iodoacetamide, 2 mM phenylmethylsulfonyl fluoride, 1 g/ml pepstatin). The membrane fractions were collected by centrifugation. AChRs were solubilized by incubating membrane fractions of the oocytes in buffer A containing 2% Triton X-100 (buffer C) at 4°C for 1 h. After removing cellular debris by centrifugation, the cleared extracts were incubated with mAb-coupled Actigel (Sterogene) or streptavidin-coupled to agarose (Sigma) (for biotinylated 2 V. Gerzanich and J. Lindstrom, manuscript in preparation. FIG. 1. AChR subunit stoichiometry and arrangement. The arrangement of subunits around the central cation channel which is known to occur in the muscle type AChRs of fish electric organs (4) is depicted in the upper left. In the case of ␣4␤2 AChRs which are known to account for most or all of the high affinity nicotine-binding sites in brain (27), the subunit stoichiometry of ␣4␤2 AChRs expressed in Xenopus oocytes using cRNAs is (␣4) 2 (␤2) 3 (10,11). It is assumed that, as in muscle AChRs (4,44), the ACh-binding sites are formed at specific interfaces between ␣ and structural subunits, which requires alternating ␣4 and ␤2 subunits around the channel. Neuronal AChRs which bind ␣-bungarotoxin are formed from ␣7, ␣8, and ␣9 subunits, and when expressed from cRNAs in Xenopus oocytes each of these subunits can form functional homomers, however, it is unknown whether additional unknown subunits might also occur in native ␣7, ␣8, or ␣9 AChRs (2,3,5). ␣7 homomers are thought to have five ligand binding sites (55). ␣3 AChRs from chick ciliary ganglion neurons appear to be a mixture with the subunit composition ␣3␤4␣5 or ␣3␤2␤4␣5 (8). It is assumed that there are two ACh-binding sites based on homology, and that these sites must be formed at specific interfaces between ␣3 and ␤ subunits. The ␣5 subunit is shown occupying the same relative position as ␤1 subunits in muscle AChRs because, like ␤1 (40, 41), ␣5 does not form ACh-binding sites when expressed alone or as a pair with other ␣ subunits (Ref. 5, and see data to follow).
AChRs) at 4°C for 6 -8 h. The resin was then washed three times with buffer A containing 0.5% Triton X-100 (buffer B), twice with buffer B containing 1 M NaCl, and again twice with buffer B. The affinitypurified AChR was eluted off the mAb-Actigel with sodium dodecyl sulfate-polyacrylamide gel electrophoresis sample buffer.
SH-SY5Y cells were cultured in a 1:1 mixture of Ham's F-12 nutrient mixture (Sigma) and Eagle's minimal essential medium (Sigma) containing nonessential amino acids. The medium was supplied with 10% fetal calf serum (HyClone). The cell monolayer was suspended with phosphate-buffered saline (PBS) containing 5 mM EDTA and pelleted in a centrifuge. The cells were homogenized and incubated in buffer C at 4°C for 1 h. Affinity purification of AChRs from SH-SY5Y cells followed the same procedure as described for AChRs from oocytes.
For immunodepletion of separate subtypes of ␣3 AChRs, oocyte extracts or cell extracts were mixed with mAb-Actigel and incubated at 4°C for 8 -10 h. The incubation time and the amount of Actigel were adjusted so that all the AChRs containing ␣5 t subunits (for oocyte extracts) or ␤2 subunits (for SH-SY5Y cell extracts), were absorbed. After depletion, the supernatant was collected by microcentrifugation, and the resin was rinsed once with an equal volume of buffer C to collect unbound AChRs. The [ 3 H]epibatidine-binding sites remaining in the supernatant were measured by solid phase radioimmunoassay (RIA) on mAb210-coated Immulon 4 (Dynatech) microwells. Nonspecific binding of ␣3 AChRs on mAb-Actigel was determined by incubating aliquots of extracts with the same amount of Actigel coupled to an irrelevant mAb (e.g. mAb306 to chick ␣7 or normal rat IgG) under the same conditions. Less than 2% of the total [ 3 H]epibatidine-binding sites were nonspecifically depleted by the mAb-Actigel.
Biotinylation of Oocyte Surface Proteins-Biotinylation was done basically as described by Zurzolo et al. (35). The oocytes were rinsed with ice-cold PBS containing 1 mM MgCl 2 , 0.1 mM CaCl 2 three times before biotinylation. The oocytes then were incubated in the same solution containing 0.5 mg/ml sulfo-NHS-biotin (Pierce) at 4°C for 30 min with mild shaking. The reaction was quenched by removing the biotin solution and incubating the oocytes in a solution of 50 mM NH 4 Cl in PBS for 10 min at 4°C, followed by rinsing the oocytes twice with PBS.
Electrophysiology-Electrophysiological recordings from oocytes injected with the various combinations of cRNAs were made as described previously by Gerzanich et al. (36).
Solid Phase RIAs-Immulon 4 (Dynatech) microtiter wells were coated with mAbs as described by Anand et al. (24). Solubilized AChRs from oocytes or SH-SY5Y cells were prepared as described above, and used directly for all assays. mAb-coated microtiter wells were incubated with Triton-solubilized AChRs (0.05-0.1 nM [ 3 H]epibatidine-binding sites) at 4°C for 8 -10 h, and then with [ 3 H]epibatidine (DuPont NEN) diluted in PBS, 0.5% Triton X-100 to different concentrations for 12-24 h to reach equilibrium. The wells were then washed three times with ice-cold PBS, 0.05% Tween 20 buffer, and the amount of radioactivity bound was determined by liquid scintillation counting. The nonspecific binding of [ 3 H]epibatidine was determined either by including 20 nM L-(Ϫ)-nicotine in the assay mixture or by processing the RIAs with lysates of noninjected oocytes. When L-(Ϫ)-[ 3 H]nicotine was used as radioligand, the incubation condition for AChRs tethered on microtiter wells was 1 h at 25°C. To compete the binding of L-(Ϫ)-[ 3 H]nicotine (20 nM) to AChRs with ACh, 1,1-dimethyl-4-phenylpiperazinium (DMPP) or cytisine, the cold ligands (in series dilution) were incubated with AChRs bound on the microtiter plates for 12-24 h at 4°C before L-(Ϫ)-[ 3 H]nicotine was added into the mixture. Saturation and competition binding data were analyzed using a nonlinear least squares curve fit method (KaleidaGraph, Abelbeck Software). The data were fit to one-site and two-site models of the Hill equation. The simpler model was accepted unless the two-site model gave a statistically better fit of the data (p Ͻ 0.05, by the F test).
Sucrose Gradient Sedimentation-Triton-solubilized AChRs from oocytes or SH-SY5Y cells were prepared as described before. Aliquots (200 l) of the lysates were layered onto 5-ml sucrose gradients (5-20% sucrose (w/w)), in 10 mM sodium phosphate buffer, pH 7.5, containing 100 mM NaCl, 1 mM NaN 3 , and 0.5% Triton X-100). The gradients were centrifuged for 1 h at 70,000 rpm in a Beckman NVT-90 rotor at 4°C. Then 11-drop fractions (about 130 l for each fraction) were collected from the bottom of the tubes and used for further analysis. If the fractions were to be analyzed by RIAs, they were collected directly in mAb210-coated wells, and incubated either with 4 nM [ 3 H]epibatidine or with 4 nM 125 I-mAb142 at 4°C for 8 -10 h. Afterwards, the wells were washed with PBS, 0.05% Tween 20 and the bound [ 3 H]epibatidine or 125 I-mAb142 were determined by liquid scintillation or ␥ counting. If the fractions were to be analyzed by Western blot, they were collected in

Functional Expression of AChR Subunit Combinations in
Xenopus Oocytes-Proper assembly and transport of AChRs to the cell surface was tested by using 125 I-mAbs to detect AChRs on the surface of intact oocytes (Fig. 2). mAb210 binds to the extracellular surface of both ␣3 and ␣5 subunits ( Fig. 2A). In order to identify only ␣5 subunits we used the reporter epitope technique described by Anand et al. (24). The reporter epitope used here is from the Torpedo ␣1 subunit, ␣395-396 (EV-IFQTPL) which can be recognized specifically by using the species-specific mAb142 as a reporter mAb (Fig. 2B). Previously, we found that C-terminal epitope tags on ␣1 subunits did not alter their function (24). Similarly, we found that AChRs containing C-terminally tagged ␣5 subunits functioned identically to those containing untagged ␣5 subunits. It is evident from Fig. 2 that both ␣3 and ␤2 (or ␣3 and ␤4) subunits are important for the proper assembly of subunits to form AChRs on the surface of oocytes. ␣5 or ␣5 t were only detected on the surface when coexpressed with ␣3 and ␤2, or ␣3 and ␤4 subunits. Neither ␣3-␣5, nor ␣5-␤2 or ␣5-␤4 subunit combinations were detected on the oocyte surface. Neither ␣5 nor ␣5 t expressed alone were detected on the oocyte surface, indicating that ␣5 subunits could not assemble as homomeric AChRs.
To further confirm the measures of surface membrane expression using 125 I-mAb binding, we used a water soluble, membrane impermeable, covalently reactive form of biotin to label the extracellular surface of AChRs expressed in oocytes followed by binding of streptavidin to detect the bound biotin (Fig. 3). Biotinylation of the ␣5 subunit further proved that it was co-assembled with ␣3 and ␤2 subunits on the surface of oocytes. The same was true for co-expression of ␣5 with ␣3 and ␤4 (data not shown).
Final evidence for expression of functional human ␣3 AChRs in Xenopus oocytes came from measurement of ion-channel properties. Only the ␣3␤2, ␣3␤4, ␣3␤2␣5, and ␣3␤4␣5 subunit combinations which were shown to be expressed on the oocyte surface gave detectable responses to ACh. Any other combination of one or two of these subunits gave no responses to ACh (data not shown). As is the case in chick (36 -38) and rat (39), human ␣3␤2 AChRs have properties distinct from those of ␣3␤4 AChRs (36).
Addition of ␣5 subunits changed the functional properties of AChRs containing ␣3␤2 or ␣3␤4 subunit combinations. ␣3␤2 AChRs desensitized more rapidly than do ␣3␤4 AChRs (Fig. 4). Addition of ␣5 subunits increased the rates of desensitization of both ␣3␤2 and ␣3␤4 containing AChRs (Fig. 4). In the case of AChRs containing ␣3 and ␤4 subunits, the addition of ␣5 subunits had little effect on either the EC 50 or efficacy of either ACh or nicotine (Fig. 5). However, in the case of AChRs containing ␣3 and ␤2 subunits, addition of ␣5 subunits increased the efficacy of nicotine from 50 to 100% and the EC 50 values for both ACh and nicotine decreased substantially, from 28 to 0.5 M in the case of ACh and from 6.8 to 1.9 M in the case of nicotine (Fig. 5). Detailed studies of the effects of ␣5 subunits on AChR function will be published separately by Gerzanich et al. 2 Subunit Composition of ␣3 AChRs-Since ␣3␤2 or ␣3␤4 subunit combinations alone could form functional AChRs, it was important to determine how efficiently ␣5 subunits assembled  lanes 2 and 3), the ␣3 and ␤2 subunits were hardly distinguishable from each other since they have similar molecular masses (57.2 kDa for ␣3, and 56.9 kDa for ␤2, as deduced from their cDNA sequences). The ␣5 subunit (53.5 kDa, as deduced from its cDNA sequence) was detected from oocytes expressing all three subunits ␣3, ␤2, and ␣5. It was purified both with mAb210-Actigel (lane 3) and streptavidin-agarose (lane 6) when it was biotinylated. The components detected by 125 I-streptavidin from uninjected oocytes (lane 1) might be oocyte surface proteins nonspecifically bound to mAb210-Actigel. They were also seen in the other two lanes (lanes 2  and 3). with these subunit pairs. This was done by measuring the fraction of the AChRs which could be bound by antibodies to epitope-tagged ␣5 subunits (␣5 t ) when ␣5 t was coexpressed with equal amounts of ␣3 and ␤2 or ␣3 and ␤4 subunits (Fig. 6). We used [ 3 H]epibatidine, a very potent agonist for ␣3 AChRs (36), to quantitate the AChRs. Under our assay conditions, ␣5 subunits were found to assemble very efficiently with ␣3 and ␤2 subunits, permitting 72% of the ␣3␤2 AChRs to be bound through the epitope tag, and permitting 55% of the ␣3␤4 AChRs to be bound through the epitope tag.
The co-assembly of ␣3, ␤2, and ␣5 subunits was further demonstrated by immunoprecipitating AChRs from crude ex-tracts of oocytes co-expressing ␣3, ␤2, and ␣5 t subunits with subunit-specific mAbs (i.e. mAb210, mAb290, or mAb142), and then detecting the precipitated subunits on Western blots using another set of subunit-specific antibodies (Fig. 7). In this case, we used polyclonal antisera against subunit-specific oligopeptides to detect ␣3 and ␤2 subunits, and mAb268 to detect ␣5 subunits. When polyclonal antiserum 3709, which is specific for the human AChR ␣3 subunit, was used on the immunoblot, we obtained a doublet band of about 57 kDa (Fig. 7A, lane 9). Using antisera specific for the human ␤2 subunit, we detected a doublet band of 55-58 kDa (Fig. 7A, lane 12). mAb268 which reacts with denatured but not native ␣5 subunits detected an FIG. 6. ␣5 subunits efficiently co-assemble with ␣3 and ␤2 subunits or ␣3 and ␤4 subunits. cRNAs for ␣3, ␤2, ␤4, ␣5, and ␣5 t subunits were injected into oocytes in varied combinations, but in equal (15 ng) amounts for each cRNA. Aliquots of the oocyte extracts were immunodepleted extensively with mAb142-Actigel, which removed all the ␣5 t -containing AChRs. By comparing [ 3 H]epibatidine-binding sites in the extracts before and after immunodepletion, the efficiency of co-assembly between ␣3␤2 and ␣5 subunits (or ␣3␤4 and ␣5 subunits) was determined. The number of [ 3 H]epibatidine-binding sites in the extract was measured by a solid phase RIA using mAb210 as the tethering mAb. Values represent the mean Ϯ S.E. from at least three separate experiments. Nonspecific depletion was assessed by substituting non-tagged ␣5 subunit for ␣5 t in coexpressing with ␣3-␤2. In this case, less than 2% of the total [ 3 H]epibatidine-binding sites were found nonspecifically absorbed by mAb142-Actigel. Triton extracts of oocytes coexpressing ␣3-␣5 t , ␤2-␣5 t , and ␤4-␣5 t were used to determine the nonspecific binding of [ 3 H]epibatidine (5 nM) in the solid phase RIA, which represented less than 1% of the total binding value.
␣5 t band at about 56 kDa (Fig. 7A, lanes 2, 5, 6, and 14). All three subunits detected on the immunoblots (Fig. 7A) have molecular masses which correspond to the expected sizes deduced from their cDNA sequences (57.2 kDa for ␣3; 56.9 kDa for ␤2; and 56.3 kDa for ␣5 t ). The doublet bands of ␣3 and ␤2 subunits seen on the Western blot might arise from variable glycosylation of the subunits in oocytes, as we had previously observed with ␣4 and ␤2 subunits (10). The fact that all the three subunits ␣3, ␤2, and ␣5 t can be co-precipitated by the ␣5 t specific mAb142 (Fig. 7A, lanes 9, 12, and 14) provided strong evidence that ␣3, ␤2, and ␣5 t were co-assembled in the oocytes. We also noticed from the immunoblot that in oocytes co-expressing ␣5 t and ␤2 subunits, aggregates of ␣5 t and ␤2 were formed, because ␣5 t was co-precipitated by the ␤2 specific mAb290 (Fig. 7A, lane 6). However, in these oocytes expressing aggregated ␣5 t ␤2 subunits, no [ 3 H]epibatidine-binding sites were detected. The aggregation of ␣5 and ␤2 subunits in oocytes was seen by sucrose gradient sedimentation as well (see Fig. 8). No ␣3-␣5 t aggregates were detected either on immunoblots (Fig. 7A, lane 10), or with the sucrose gradient sedimentation assay. Lacking antibodies specific for the human ␤4 subunit, we could not apply the same experiments described above to demonstrate the co-assembly of ␣3, ␤4, and ␣5 subunits; but such AChR complexes were formed in the oocytes (see Fig. 6).
Native ␣3 AChRs from the human neuroblastoma cell line SH-SY5Y were studied by a similar approach. We found that both ␣3 and ␣5 subunits could be co-purified by the ␤2-specific mAb290 coupled to Actigel (Fig. 7B, lanes 2 and 3), which implied the co-assembly of all three subunits in one pentamer. The immunopurified AChRs could be a mixture of ␣3␤2 AChRs, ␣3␤2␣5 AChRs, and ␣3␤2␤4␣5 AChRs. We determined the fraction of ␣3 AChRs in SH-SY5Y cells which contained ␤2 subunits by measuring the fraction of [ 3 H]epibatidine-labeled ␣3 AChRs which could be bound by the ␤2-specific mAb290 coupled to Actigel. We found that at least 56% of the total ␣3 AChRs contained ␤2 subunits. This is a much larger fraction than the 20% of ␣3 AChRs in chick ciliary ganglion which contain ␤2 subunits (8). Here we did not directly study the fraction of AChRs which contained ␤4 subunits because no human ␤4 subunit-specific antibodies were available.
Sucrose Gradient Sedimentation of ␣3 AChRs--We studied the sedimentation behavior of human ␣3 AChRs expressed in Xenopus oocytes. By comparing their sedimentation properties with those of native ␣3 AChRs from SH-SY5Y and those of chick ␣4␤2 AChRs expressed in oocytes, we found that in all cases the functional complexes (indicated by the binding peak for [ 3 H]epibatidine) co-sedimented with native ␣3 AChRs in the 11 S region (Fig. 8). We suggest that this 11 S component corresponds to fully assembled pentamers. By Western blot analysis, as well as solid phase binding assay of the fractions from sucrose gradient sedimentation with subunit specific mAbs, we were able to detect the majority of both ␣3 and ␣5 subunits in the 11 S region of the gradient, the same region where the functional peak of the AChRs were located (Fig. 9). The results in Fig. 9 imply efficient co-assembly of ␣3, ␤2, and ␣5 subunits because most or all of the subunits were assembled into complexes the size of native AChRs.
Sucrose gradient analysis (Fig. 8) further confirmed data from other types of experiments (Fig. 7) which indicated that ␣5 subunits will associate with ␤2 subunits if that pair of subunits is co-expressed, but that ␣5 subunits will not associate with ␣3 if that pair is co-expressed. It may be that ␣5 cannot associate with ␣3 until ␣3 has associated with ␤2. Alternatively, whatever associations are made are not stable during solubilization in Triton X-100. Similarly, it has been shown with muscle AChRs that ␣1 efficiently assembles with ␥ and ␦ subunits at interfaces which permit the formation of ACh-binding sites, but that ␤1 does not efficiently associate with ␣1 subunits in complexes stable in Triton X-100 until ␣1 has associated with ␥ and ␦ subunits (40 -43 3 and 7), ␣5 t ␤2 (lane 6), ␣3␣5 t (lane 10), and ␣3␤2␣5 t (lanes 2, 5, 9, 12, and 14) were expressed in oocytes and solubilized in Triton X-100 solution. mAb-Actigels were used to immunopurify AChRs from the oocyte extracts: mAb210 to ␣3 and ␣5 subunits (lanes 1-3); mAb290 to ␤2 (lanes 4 -7); and mAb142 to ␣5 t (lanes 8 -14). Rabbit antisera were used to detect ␣3 and ␤2 subunits on the immunoblots (lanes 8 -12). mAb268 was used to detect ␣5 t subunits (lanes 1-7, 13, and 14). The signals on the blots were specific, since they were not detected when Triton X-100 extracts from uninjected oocytes were treated in the same procedure (lanes 1, 4, 8, 11, and 13). Panel B, Triton X-100 extracts from the human neuroblastoma cell line SH-SY5Y were immunopurified with mAb290-Actigel (specific for ␤2 subunits). The immunoblots were probed with rabbit anti-human ␣3 serum and mAb268 (to ␣5 subunits). As a negative control, the immunoblot was also probed only with 125 I-goat anti-rabbit IgG. The molecular mass markers are as described in the legend to Fig. 3. report from this laboratory (36), epibatidine, an azabicycloheptane alkaloid from the skin of an Ecuadoran frog (39), was successfully employed as a potent nicotinic agonist in characterizing AChRs and shown to provide a useful label for ␣3 AChRs. In this study, [ 3 H]epibatidine was used to study the pharmacological properties in solid phase RIAs of human ␣3 AChRs expressed in Xenopus oocytes. For saturation binding, ␣3 AChRs in different subunit combinations, i.e. ␣3␤2, ␣3␤2␣5 t , ␣3␤4, and ␣3␤4␣5 t , were tethered on mAb-coated microwells.
Switching from ␤2 to ␤4 subunits has a large effect (41-fold) on the affinity of ␣3 AChRs for epibatidine, as shown in Fig. 10. This result is consistent with our previous study of human ␣3␤2 AChRs and ␣3␤4 AChRs in Xenopus oocytes with voltage clamp analysis, which demonstrated that ␣3␤2 AChRs have higher binding affinity for epibatidine than do ␣3␤4 AChRs (36). The Hill coefficients (nH) of [ 3 H]epibatidine saturation binding curves for all four ␣3 AChRs tested were close to 1.0 (Table I), suggesting that there was only one class of [ 3 H]epibatidine-binding sites in each subunit combination.
Addition of ␣5 subunits to ␣3 and ␤2, or ␣3 and ␤4 subunits has little (2-fold or less) effect on the affinities of these AChRs for epibatidine, as shown in Fig. 10. If ␣5 subunits assemble as indicated in Fig. 1 at the position of ␤1 subunits, and thus do not form ACh binding interfaces with ␣3 or ␤ subunits, one FIG. 8. Sucrose gradient sedimentation analysis of human ␣3 AChRs. Human ␣3 AChRs expressed in oocytes or SH-SY5Y cells were sedimented on 5-20% sucrose gradients. Fractions are numbered from the bottom of the gradients. ␣3 AChRs were quantitated by [ 3 H]epibatidine binding (4 nM) in solid phase RIAs on mAb210-coated microwells (Panels A-C). As size standards for AChR pentamers, AChRs from Torpedo electric organ and chick ␣4␤2 AChRs expressed in oocytes were also sedimented on parallel gradients. Torpedo AChRs were tethered on mAb210-coated wells and labeled with 125 I-␣-bungarotoxin (3.5 nM). Chick ␣4␤2 AChRs were tethered on wells coated with mAb299 specific for the ␣4 subunit and labeled with [ 3 H]epibatidine (4 nM). In Panels D, E, and F, ␣5 subunits in each fraction were quantitated by immunoblot analysis using 125 I-mAb268 as probe, followed by optical densitometry of the signals on the blot. The optical densitometry profile is aligned with fractions of parallel gradients of ␣3 AChRs quantitated with [ 3 H]epibatidine binding shown in Panels A-C. Co-expression of ␣3, ␤2, and ␣5 subunits results in fully assembled AChRs which can bind epibatidine (Panel C), whereas co-expression of ␤2 and ␣5 results in assembly of a wide array of aggregates which cannot bind epibatidine (Panel D). Co-expression of ␣3 and ␣5, or expression of ␣5 alone results in only unassembled ␣5 subunits (Panels E and F). 9. Efficient assembly of ␣3 t , ␤2 and ␣5 t subunits in oocytes demonstrated by sucrose gradient sedimentation. Human AChR subunits ␣3 t (␣3 subunit tagged with the mAb236 epitope), ␤2 and ␣5 t (␣5 subunits tagged with the mAb142 epitope) were co-expressed in oocytes, solubilized using Triton X-100, and sedimented on 5-20% sucrose gradients. Fractions are numbered from the bottom of the gradients. Panel A, properly assembled ␣3 AChRs were quantitated by [ 3 H]epibatidine (4 nM) on mAb210-coated microwells. Panel B, ␣3 t subunits in each fraction were detected by immunoblot assay with 125 I-mAb236 as probe. Signals on the immunoblot were quantitated by optical densitometry. Panel C, ␣5 t subunits in each fraction were quantitated by 125 I-mAb142 (4 nM) binding on mAb210-coated microwells. Panel D, denatured ␣5 t subunits were also identified by 125 I-mAb268 to ␣5 on an immunoblot of fractions from the gradient. might expect that the effects of ␣5 on the ligand binding properties of AChRs of which it is a part would be rather small, as we have observed.
In addition to [ 3 H]epibatidine, the pharmacology of heterologously expressed human ␣3␤2 AChRs and ␣3 AChRs from SH-SY5Y has also been studied using L-(Ϫ)-[ 3 H]nicotine, and with ACh, cytisine, and DMPP to compete for the binding of  Table II. Both the cloned ␣3␤2 AChR subtype and the mixture of ␣3 subunit containing AChRs in SH-SY5Y cells exhibit similar affinities for cytisine and DMPP and only a 2-fold difference in affinity for ACh. However, nicotine, like epibatidine, distinguishes substantially in affinity between the mixture of subtypes present in the neuroblastoma.

DISCUSSION
In this study, we characterized human ␣3 AChRs heterologously expressed in Xenopus oocytes and compared them with native ␣3 AChRs isolated from the human neuroblastoma cell line SH-SY5Y. Our main findings were: 1) human AChR ␣5 subunits can co-assemble efficiently with ␣3/␤2 and ␣3/␤4 subunit combinations to form functional AChRs on the cell surface when they are co-expressed in oocytes, but ␣5 subunits expressed alone or in paired combination with ␣3, ␤2, or ␤4 subunits are not expressed on the cell surface; 2) ␣5 subunits are not assembled as homomers, nor do they assemble as pairs with ␣3 subunits; but they will form intracellular complexes with ␤2 subunits which do not contain ACh-binding sites; 3) as structural subunits, ␤2 and ␤4 make different contributions to the pharmacological, as well as ion-channel properties of human ␣3 AChRs; 4) the co-assembly of ␣5 subunits increases the rate of desensitization of both AChRs containing ␣3 and ␤2 or ␣3 and ␤4 subunits; 5) the co-assembly of ␣5 subunits with ␣3 and ␤2 subunits substantially reduces the already rather low EC 50 values for ACh and nicotine while increasing the efficacy of nicotine from 50 to 100%, yet in the case of ␣3 and ␤4 H]epibatidine to human ␣3 was done in mAb210-coated microwells (for ␣3␤2 AChRs and ␣3␤4 AChRs) or mAb142-coated microwells (for ␣3␤2␣5 t AChRs and ␣3␤4␣5 t AChRs) at 4°C for 24 h to reach equilibrium. Nonspecific binding was determined by using Triton X-100 extracts of uninjected oocytes and subtracted from each data point. The binding curves were generated by using the least squares curve fit method (KaleidaGraph). Scatchard analysis of the data were shown in the insets. The means Ϯ S.E. (nϾ3) for the K D values were summarized in Table I.  subunits the co-assembly with ␣5 subunits has little effect on the relatively high EC 50 values for ACh and nicotine and nicotine remains a full agonist; 6) the co-assembly of ␣5 subunits with both ␣3 and ␤2 or ␣3 and ␤4 in oocytes does not dramatically change the binding affinities of the resulting AChRs toward [ 3 H]epibatidine; 7) native ␣3 AChRs from SH-SY5Y cells have ␣5 subunits associated, and about half of these AChRs have ␤2 subunits; and 8) binding of epibatidine and nicotine to ␣3 AChRs from SH-SY5Y distinguishes subtypes with different binding affinities, and those with ␤2 subunits, like cloned ␤2 containing ␣3 AChRs, have higher affinity for epibatidine; 9) the mixture of native ␣3 AChR subtypes in SH-SY5Y cells have similar affinities to cloned ␣3␤2 AChRs for ACh, cytisine, and DMPP.
It has been reported previously that ␣3 AChRs from chick ganglia can have as many as four different subunits (i.e. ␣3, ␤2, ␤4, and ␣5) in one pentamer (8). Here we demonstrated, by immunoprecipitation and Western blot analysis using various subunit-specific mAbs and polyclonal antisera, that human AChR ␣3, ␤2, and ␣5 subunits were co-assembled in oocytes. We found that tagging of ␣5 subunits with a reporter epitope (24) at the C-terminal extracellular end was a very useful tool for the isolation and detection of the AChRs containing that subunit. In addition to the investigation of subunit composition, reporter epitope-tagged ␣5 and the reporter mAb142 were also very useful for sorting out ␣3␤2␣5 t AChRs from a mixture of ␣3␤2 and ␣3␤2␣5 t in solid phase RIAs. With the ␣5 subunit tagged, we were able to assay the binding properties of ␣3␤4␣5 t AChRs with [ 3 H]epibatidine as well. The fact that tethered ␣5 t -containing AChR from oocytes co-expressing ␣3, ␤4, and ␣5 t could bind [ 3 H]epibatidine proved indirectly that ␣3, ␤4, and ␣5 t subunits were in one pentamer. We did not have direct evidence that ␣3, ␤4, and ␣5 t were in one complex by immunopurification and Western blot analysis because no antibodies were available to detect human ␤4 on Western blots, which also limited our study of AChRs containing all four different subunits, ␣3, ␤2, ␤4, and ␣5.
It is important to determine the functional effects of ␣5 associating with ␣3 containing AChRs because most or all of the ganglionic type ␣3 AChRs studied in chick ganglia (8) or human neuroblastomas like SH-SY5Y have ␣5 subunits associated with them. We found that the effects of ␣5 subunits depended on whether they associated with ␣3␤2 or ␣3␤4 subunit combinations. ␣5 increased the rate of desensitization with either combination of subunits. The most obvious effects of ␣5 were on the ␣3␤2 combination, where EC 50 values for ACh and nicotine were reduced to lower concentrations and nicotine was shifted from a partial to a full agonist. The association of ␣5 with various AChR subtypes might also alter assembly, turnover, transport, or regulatory properties which we have not assayed but which might be important in vivo. In chick brain about half of the AChRs with high affinity for nicotine are composed of ␣4 and ␤2 subunits and in mammalian brain ␣4␤2 AChRs account for at least 90% of the high affinity nicotine binding (27). In chick brain a small fraction of ␣4␤2 AChRs also have ␣5 subunits associated (29). It was recently reported that chick ␣4␤2␣5 AChRs expressed in oocytes have about 125-fold lower EC 50 for ACh and twice the conductance of ␣4␤2 AChRs (18).
Sucrose gradient sedimentation and Western blot analysis of expressed pairs of ␣3-␤2, ␣3-␣5, or ␤2-␣5 subunits in oocytes indicated that ␣3 and ␤2, as well as ␤2 and ␣5 can associate with each other as a pair, but that ␣3 and ␣5 subunits cannot associate as a pair, at least in Triton X-100. This observation hints at a putative order of assembly of the component subunits. ␣3 clearly assembles efficiently with ␤2 or ␤4 to form functional AChRs. It may be that ␣5 can only assemble with ␣3 subunits after they have associated with ␤ subunits. In this respect ␣5 may resemble ␤1 subunits which do not assemble efficiently with ␣1 in complexes stable in Triton X-100 unless ␣1 is associated with ␥ or ␦ (40, 41). ACh-binding sites are thought to be formed at the interfaces between ␣1 and ␥ or ␦ subunits (3,4,44). ACh binding properties of ␣3 AChRs can vary greatly depending on the presence of ␤2 or ␤4 subunits (9), as we have also observed here, indicating that ␣3␤ interfaces form ACh-binding sites. For example, we found that ␣3␤2 AChRs and ␣3␤4 AChRs differed 41-fold in affinity for epibatidine. Co-assembly with ␣5 subunits did not alter these binding affinities, suggesting that coassembly with ␣5 does not displace the ␣3␤ interfaces which form the ligand binding sites. Because of homology considerations and their observed similar sizes on sucrose gradients, it is compelling to think that ␣3 AChRs are pentamers. In order to account for ␣3␤2 AChRs or ␣3␤4 AChRs with ligand binding properties that are not greatly altered by the presence of ␣5, it seems likely that ␣5 occupies the position depicted in Fig. 1, where it interacts with the non-ligand binding interfaces of ␤ and ␣3 subunits. We have shown that, wherever ␣5 associates, it can alter EC 50 values, efficacy, and desensitization rates. It may be that this is because these effects depend on cooperative conformation changes effecting many parts of the AChR in addition to the subunit interfaces which form the ligand binding sites.
Considering our results of radioimmunoassay and immunoblot analysis, as well as the Northern analysis reported previously (20), we suggested that the ␣3 AChRs in SH-SY5Y cells can be sorted into at least two groups. One group includes AChRs with ␤2 subunits (potentially some combination of ␣3␤2, ␣3␤2␣5, ␣3␤2␤4, and ␣3␤2␤4␣5) which have higher apparent affinity for [ 3 H]epibatidine. Another group has ␣3 AChRs without ␤2 subunits (i.e. ␣3␤4 and ␣3␤4␣5) with lower apparent affinity for [ 3 H]epibatidine. Because of the limited resolution of the saturation binding assay in Fig. 11, we were not able to distinguish AChRs in the same group (with different subunit combinations) according to their difference in binding affinity for [ 3 H]epibatidine. mAb210, a mAb made to the main immunogenic region on ␣1 subunits, was effectively used in our study of human ␣3 AChRs, because it can react with native human ␣3 and ␣5 subunits. Another mAb to the main immunogenic region, mAb35 has the same properties (data not shown) and has been used in similar studies on chick ␣3 AChRs (8). Comparing the sequences of main immunogenic region amino acids (corresponding to ␣1 66 -76) (2), we found high homology among human ␣1, ␣3, and ␣5 subunits (and the closely related ␤3 subunit), which explains the cross-reaction of mAb210 with human ␣3 and ␣5. Since mAb210 has the same specificity as the majority of anti-AChR autoantibodies from myasthenia gravis patients (45), it seems likely that some myasthenia gravis autoantibodies could also cross-react with human ␣3 and ␣5. Since ␣3, ␤4, and ␣5 RNAs (46) as well as ␣1 RNA (47)(48)(49) have been identified in thymus, it may be that neuronal AChRs in the thymus (possibly ␣3 AChRs) as well as ␣1 AChRs may also be involved in induction of the autoimmune response in myasthenia gravis, or in its secondary pathological effects (50). On the other hand, the main target of the autoimmune attack in myasthenia gravis is muscle ␣1 AChRs (51), and myasthenia gravis has only on rare occasions been associated with peripheral (52) or central nervous system anomalies (53,54). Myasthenia gravis patient antibodies were not found to bind to high affinity nicotine binding (␣4␤2) AChRs or to ␣-bungarotoxin binding (␣7) AChRs from human brain (54), but these studies would not have detected interactions with ␣3 AChRs of the types studied here.