The role of the cystine loop in acetylcholine receptor assembly.

Nicotinic acetylcholine receptors (AChRs) are composed of alpha, beta, gamma, and delta subunits, assembled into alpha2betagammadelta pentamers. A highly conserved feature of ionotropic neurotransmitter receptors, such as AChRs, is a 15-amino acid cystine "loop." We find that an intact cystine loop is necessary for complete AChR assembly. By preventing formation of the loop with 5 mM dithiothreitol, AChR subunits assemble into alphabetagamma trimers, but the subsequent steps in assembly are blocked. When alpha subunit loop cysteines are mutated to serines, assembly is blocked at the same step as with dithiothreitol. In contrast, when beta subunit loop cysteines are mutated to serines, assembly is blocked at a later step, i.e. after assembly of alphabetagammadelta tetramers and before the addition of the second alpha subunit. After formation of the cystine loop, the alpha subunit undergoes a conformational change, which buries the loop. This conformational change is concurrent with the step in assembly blocked by removal of the disulfide bond of the cystine loop, i.e. after assembly of alphabetagamma trimers and before the addition of the delta subunit. The data indicate that the alpha subunit conformational change involving the cystine loop is key to a series of folding events that allow the addition of unassembled subunits.

The molecular events involved in subunit folding and assembly of large oligomeric proteins remain largely uncharacterized. The subunit folding and oligomerization events that take place during the assembly of ion channels are particularly complex, since the finished product requires the correct oligomeric arrangement and subunit stoichiometry for proper function (1). In terms of their structure, the best characterized ion channels are the muscle-type nicotinic acetylcholine receptors (AChRs), 1 which are the neurotransmitter receptors responsible for rapid signaling between motor neurons and skeletal muscle. Muscle-type AChRs are composed of four distinct, homologous subunits, ␣, ␤, ␥, and ␦, which assemble into pentamers with the subunit stoichiometry of ␣ 2 ␤␥␦.
Although AChR assembly is a slow process that takes ϳ2 h to complete (2), assembly intermediates have been difficult to isolate. A number of laboratories turned to expression of less than the full complement of subunits in different heterologous expression systems to isolate assembly intermediates (3)(4)(5)(6).
Based on their findings, the "heterodimer" model was proposed, where the ␣ subunit must first fold or "mature," as assayed by the formation of the ␣-bungarotoxin (BuTx) binding site and antigenic epitopes, before assembling with other subunits. The mature ␣ subunit assembles with ␥ or ␦ subunits in parallel to form ␣␥ and ␣␦ heterodimers, and the heterodimers associate together and with ␤ subunits to form ␣ 2 ␤␥␦ pentamers.
We have developed techniques that have allowed isolation of assembly intermediates in cells stably expressing all four AChR subunits (7,8) and have obtained results at odds with the heterodimer model. Instead of heterodimers, two partially assembled complexes, ␣␤␥ trimers and ␣␤␥␦ tetramers, were isolated. ␣␤␥ trimers, which assemble extremely rapidly, were assembled first into ␣␤␥␦ tetramers and then into ␣ 2 ␤␥␦ pentamers. Our data demonstrated that assembly occurs sequentially, each step being the addition of an uncomplexed subunit. We also demonstrated that the ␣ subunit maturation steps, which were thought to precede its assembly, occurred after assembly into ␣␤␥ trimers but prior to the addition of the ␦ subunit. These folding events require a specific combination of subunits and correlate in time with the ␦ subunit addition. The data led us to suggest that the ␣ subunit maturation steps are folding events forming the ␦ subunit recognition site, i.e. the site where the ␦ subunit associates with the ␣␤␥ trimer.
Our goal has been to identify posttranslational processing sites and regions on the AChR subunits involved in AChR subunit folding and assembly. A good candidate is the highly conserved region defined by a pair of cysteine residues separated by a stretch of 13 amino acids, which is found on the neurotransmitter-binding, extracellular domain of the subunits (Fig. 1, A and B). The two cysteines form a disulfide bridge on all four Torpedo AChR subunits (9,10). Analogous cystine "loops" appear to form on other neurotransmitter-gated ion channel subunits, which include all muscle and neuronal AChR subunits and all GABA A , glycine, and 5HT 3 receptor subunits. Other residues in the loop are identically conserved across species from Caenorhabditis elegans to mammals and are even conserved on some of the glutamate receptor subunits (11). Site-directed mutations of the cysteines in the ␣ subunit prevented BuTx binding site formation and reduced AChR expression (12), which suggested that the cystine loop is critical for subunit conformational stability or assembly. A recent study where the conserved proline in the cystine loop (see Fig.  1, A and B) was mutated also suggested that the cystine loop may play a role in subunit assembly (13). However, mutated ␣ or ␤ subunits lacking the cystine loop disulfide bond assembled with other wild type subunits (14,15) and produced functional receptors (15). Based on these data, it was suggested that formation of the cystine loop is not required for subunit assembly but instead plays an important role in the rate that subunits are degraded and in the efficiency of transport of AChRs to the cell surface. Contrary to this viewpoint, we demonstrate in this paper that an intact cystine loop is essential for proper AChR subunit assembly. Elimination of the cystine loop separately on the ␣ and ␤ subunits blocks different steps during assembly. For the ␣ subunit, the cystine loop undergoes a conformational change, which appears to be an event required for assembly to continue.

Cystine Loop Mutations, Cell Lines, and AChR Subunit Expression-
The cystine loop mutations, ␣ 128 , ␣ 142 , and ␤ 128 (15), where the subscript indicates which cystine residues were replaced by serine residues, were a generous gift from Dr. Sumikawa. The cystine loop mutation, ␣ 128/142 , was created by inserting the ␣ 142 BclI fragment back into ␣ 128 . ␣ 128/142 and ␤ 128 were subcloned into the EcoRI site of pSVDF4 (16). The ␣ 128/142 ␤␥␦ and ␣␤ 128 ␥␦ cell lines were established by stably transfecting each of the two cystine loop mutations along with the appropriate wild type subunit constructs plus the thymidine kinase gene (tk) into mouse fibroblast L cells deficient in tk as described previously (7). To establish the ␣ 128 cell line, mouse NIH3T3 cells were infected with the retroviral recombinant, pDOJ, which contained the ␣ 128 subunit cDNA in the EcoRI site as well as the neomycin resistance gene. The ␣␤␥␦ and ␣␤␥ cell lines, which stably express the indicated Torpedo AChR subunits, were previously described (7,8).
The assembly of the Torpedo subunits is temperature-dependent (7,17). To allow assembly to occur, the temperature was dropped from 37 to 20°C. The transfected Torpedo subunit cDNAs in the ␣␤␥␦ cell line are under the control of SV40 promoters (7). To enhance expression of Torpedo AChR subunits, the medium (Dulbecco's modified Eagle's medium (DMEM; JRH Scientific) plus 10% calf serum (Hyclone) and HAT (15 g/ml hypoxanthine, 1 g/ml aminopterin, and 5 g/ml thymidine), was replaced with fresh DMEM supplemented with 20 mM sodium butyrate (NB medium; Baker) 36 -48 h prior to the experiment.
Metabolic Labeling and Immunoprecipitations-To pulse-label the subunits, 10-cm cultures were first washed with PBS and starved of methionine for 15 min in methionine-free DMEM. Cultures were incubated at 5% CO 2 and labeled in 2 ml of methionine-free DMEM, sup-plemented with 20 mM sodium butyrate and 333 Ci of [ 35 S]methionine for 30 min at 37°C. The labeling was stopped with the addition of DMEM containing 5 mM methionine and, if not "chased," two washes of ice-cold PBS. To follow the subsequent changes in the labeled subunits, the cells were chased. Specifically, these cells were washed two more times with DMEM containing 5 mM methionine and incubated for various times in NB medium at 20°C in the absence or presence of 5 mM dithiothreitol (DTT). After a [ 35 S]methionine pulse chase, cultures were rinsed with ice-cold PBS, scraped, and pelleted by centrifugation at 5,000 ϫ g for 3 min, and the pellets were resuspended in lysis buffer: 150 mM NaCl, 5 mM EDTA, 50 mM Tris, pH 7.4, 0.02% NaN 3 , 2 mM phenylmethylsulfonyl fluoride, 2 mM N-ethylmaleimide plus 1% solubilizing agent. In some of the experiments, 1 mM MgATP replaced the 5 mM EDTA to reduce the amount of actin that nonspecifically precipitated with the subunits. To solubilize the labeled subunits, a Lubrolphosphatidylcholine mixture composed of 1.83 mg/ml phosphatidylcholine and 1% Lubrol was used. Antibody-subunit complexes were precipitated with Protein G-Sepharose and electrophoresed on 7.5% SDS-polyacrylamide gels, fixed, enhanced for 30 min, dried on a gel dryer, and exposed to film at Ϫ70°C with an intensifying screen. Autoradiographs were quantified by scanning densitometry using a flatbed scanner and analyzed with the Intelligent Quantifier software from BioImage. 125 I-BuTx Binding-To measure cell surface 125 I-BuTx binding, cultures were grown at 37°C for 36 h in NB medium and then shifted to 20°C for 48 h. Cultures were washed with PBS and incubated at room temperature in PBS containing 4 nM 125 I-BuTx (140 -170 cpm/fmol) for 2 h, which results in the saturation of binding. Cultures were washed again and solubilized, and the cell surface counts were determined by ␥-counting, or counts were immunoprecipitated with appropriate antibodies and counted. Total cell 125 I-BuTx binding was measured after solubilization. In this case, cell lysates were incubated in 10 nM 125 I-BuTx at 4°C overnight to saturate binding at this lower temperature. The 125 I-BuTx in the lysates was then immunoprecipitated with the indicated antibodies.
Sucrose Gradients-To distinguish subunit complexes on the basis of their size, solubilized subunits (labeled with [ 35 S]methionine and/or bound by 125 I-BuTx) were sedimented on sucrose gradients. For this procedure, lysates were layered on a 5-ml 5-20% linear sucrose gradient prepared in the appropriate lysis buffer. Gradients were centrifuged in a Beckman SW 50.1 rotor at 40,000 rpm to 2 t ϭ 9.0 ϫ 10 11 . ϳ17 fractions (300 l) were collected from the top of the gradient, and the appropriate antibodies were then added to the fractions to be assayed.

Reduction of Disulfide Bonds Alters Subunit Assembly-
When added extracellularly to cultured cells, DTT reaches the endoplasmic reticulum and prevents the formation of protein disulfide bonds without altering most other cellular functions (18). Cells stably expressing the four Torpedo AChR subunits were subjected to an [ 35 S]methionine pulse-chase protocol in the absence or presence of 5 mM DTT (Fig. 2, A and B). Labeled subunits were immunoprecipitated with either a ␥ subunitspecific monoclonal antibody (mAb 168) or ␣ subunit, conformation-dependent mAb 14, to assay for subunit assembly and formation of the mAb 14 epitope. During a 30-min pulse of [ 35 S]methionine, ␣␤␥ trimers formed as shown by the coprecipitation of predominantly ␣ and ␤ subunits with the ␥ subunits ( Fig. 2A). During the chase in the absence of DTT, progressively more ␦ subunits are added to the trimers, followed by the addition of the second ␣ subunit as shown by the doubling in the amount of ␣ subunit relative to the other coprecipitated subunits. These two subunit additions are better resolved by the mAb 14 immunoprecipitations (Fig. 2, A and C).
Since the mAb 14 epitope forms on ␣␤␥ trimers just prior to the addition of the ␦ subunit (8), these immunoprecipitations show the complete time course of both the addition of the ␦ and second ␣ subunits to the ␣␤␥ trimers.
In the presence of DTT, the subunits clearly retain the ability to assemble into ␣␤␥ trimers and with approximately the same efficiency as occurs in the absence of DTT (Fig. 2B). This result is at odds with a recent study (19), which suggested that 5 mM DTT completely blocks AChR subunit assembly. There  (30), the rat brain nAChR ␣2, ␤2 (31) and ␣7 (32) subunits, the C. elegans nAChR subunit deg-3 (33), the calf brain GABA A receptor ␣1 and ␤1 subunits (34,35), the rat brain glycine receptor ␣1 subunit (36), the mouse 5HT3 receptor subunit (37), and the rat brain glutamate receptor subunit Glu R1 (11). B, the cystine loop is located within the subunit's N-terminal, extracellular region. The putative secondary structure of AChR subunits is displayed, showing the four membranespanning regions, M1-M4 (for a recent review, see Ref. 38).
are several differences between this and the other study that explain the conflicting results. Probably the most important difference in terms of the ability to observe ␣␤␥ trimers was that their only assay for measuring the assembly of ␣ subunits with other subunits was an immunoprecipitation with ␦ subunit-specific antibodies. Obviously, ␣␤␥ trimers could never be observed with this assay. Another difference was the protocol used to solubilize the AChR subunit complexes. We have shown previously that solubilization in 1% Triton X-100 causes the assembly intermediates to dissociate and that the subunit associations are stable when solubilized with a combination of Lubrol PX and phosphatidylcholine (8).

FIG. 2. Reduction of disulfide bonds blocks AChR subunit assembly.
A and B, AChR subunit assembly in the absence or presence of 5 mM DTT. A mouse L cell line, stably expressing all four Torpedo AChR subunits (the ␣␤␥␦ cell line; Ref. 7), was pulse-labeled with [ 35 S]methionine for 30 min at 37°C and chased at 20°C in the absence or presence of 5 mM DTT for the indicated times. Labeled subunits were immunoprecipitated with either the ␥ subunit-specific mAb 168 or the ␣ subunit, conformation-dependent mAb 14 (24). The band labeled as ␣Ј has previously been shown to be different from the ␦ subunit (8) (see also Fig. 5B), although it migrates just above the ␦ subunit. Further evidence that this band is not the ␦ subunit is displayed in lane 1 of Fig. 2A. Cells stably expressing only the Torpedo ␣, ␤, and ␥ subunits were subjected to the same pulse-chase protocol as the ␣␤␥␦ cells and immunoprecipitated with mAb 168. The ␣Ј band coprecipitates with the other subunits in the absence of any ␦ subunit expression. C, time course of mAb 14 epitope formation. Displayed is the quantification of the mAb 14-precipitated ␣, ␤, ␥, and ␦ subunit bands analyzed by SDS-polyacrylamide gel electrophoresis and quantified from resultant fluorographs by scanning densitometry. Also displayed for comparison are the scanned values of the ␦ subunits coprecipitating with the precipitated ␥ subunits. The percentage assembled values are the scanned values divided by the value for the ␤ subunit at the 48-h chase time (100%) and are shown to emphasize the time course of the ␣ subunit doubling. D, the rate of ␥ subunit degradation in the absence or presence of DTT. Displayed are the scanned values for the ␥ subunit bands chased in the absence or presence of DTT and precipitated by the ␥ subunit-specific mAb (Fig. 2, A and B). Also displayed are values for the ␤ subunits that coprecipitate with the ␥ subunits in the absence of DTT ( Fig. 2A). The data are plotted on a semilog scale. In the absence of DTT, the decay of the ␥ subunit signal is biphasic. The slowly decaying component corresponds to assembled ␥ subunits as shown by the similar rate of decay of the assembled ␤ subunits. To estimate the rate of decay of unassembled ␥ subunits, the 48-h value was subtracted from the other values and plotted as the ␥ adjusted symbols. E, the presence of DTT blocks BuTx site formation. 125 I-BuTx binding to the cell lysate of ␣␤␥␦ cells was measured for cells grown in the absence or presence of 5 mM DTT for different lengths of time. The 0 time point is the time at which the cultures were shifted from 37 to 20°C to start subunit assembly. A single 10-cm culture was used for each time point.
After ␣␤␥ trimers assemble in the presence of 5 mM DTT, the ensuing subunit associations, the addition of the ␦ and second ␣ subunits, fail to occur, and the ␣␤␥ trimers that had assembled were degraded 48 h after the [ 35 S]methionine pulse (Fig.  2B). ␥ subunits in the presence of DTT degrade at about the same rate as unassembled ␥ subunits in the absence of DTT (Fig. 2D). Subunit complexes degrade more rapidly in the presence of DTT because the complexes are no longer stabilized by events subsequent to trimer formation. Although there are almost as many ␣␤␥ trimers present with DTT at the times when ␦ subunits normally assemble with ␣␤␥ trimers (Fig. 2, A  and B; 3-and 6-h chase), the formation of ␣␤␥␦ tetramers and ␣ 2 ␤␥␦ pentamers is inhibited in the presence of DTT. Furthermore, the BuTx binding site (Figs. 2E and 3A) and the mAb 14 epitope (Fig. 2B) do not form on the ␣␤␥ trimers in the presence of DTT. We conclude that the addition of DTT blocks subunit assembly after the association of the ␣, ␤, and ␥ subunits and before the BuTx binding site and mAb 14 epitope form on the ␣␤␥ trimers. Since formation of the BuTx binding site and mAb 14 epitope precede the addition of the ␦ subunit (8), the data suggest that a block of these folding events by DTT prevents subsequent subunit associations.
Elimination of Cystine Loop Disulfide Bond on the ␣ Subunit Blocks Assembly-The addition of DTT in the above experiments prevents disulfide bond formation for all AChR subunits as well as for other proteins that might affect AChR subunit assembly. To test whether the cystine loop on AChR subunits is involved in the DTT block of assembly, we obtained mutations of the Torpedo AChR ␣ and ␤ subunits in which cysteines forming the cystine loop were replaced by serines (15). Of the four AChR subunits, only ␣ subunits contain an additional cystine formed between adjacent cysteines (cysteines 192 and 193) located at the ACh binding site (20) (Fig. 1B). Deletion mutations eliminating that cystine cause the loss of ACh binding but have no effect on BuTx binding, subunit assembly, or cell surface expression (12,14). An ␣ subunit construct was created, ␣ 128/142 , where both cystine loop cysteines were re- ]methionine and chased for the indicated times. Lane 1, labeled subunits were immunoprecipitated with a mixture (cocktail) of ␣, ␤, ␥, and ␦ subunit-specific antibodies, which consisted of ␣ and ␤ subunit-specific polyclonal antiserum (39) and the ␥ and ␦ subunit-specific mAb 88b (American Tissue Culture Collection). Lanes 2-6, labeled subunits were immunoprecipitated with the ␣ subunit-specific polyclonal antiserum, eluted from protein G-Sepharose with a 0.1 M glycine buffer (pH 2.5), and precipitated a second time with the ␣ subunit-specific polyclonal antiserum. Only significant amounts of the ␤ and ␥ subunits coprecipitate with the ␣ 128/142 subunit, and the efficiency of these interactions was reduced 2-3-fold relative to wild type subunit assembly (see Fig. 2A, lane 2, and Fig. 6C, lane 1). B, formation of the mAb 35 epitope during subunit assembly in the ␣ 128/ 142␤␥␦ cell line. ␣ 128/142 ␤␥␦ cells were pulse-labeled and chased for the same times as in Fig. 4A (lanes 2-6) except that labeled subunits were immunoprecipitated with mAb 35. C, the rate of ␣ 128/142 subunit degradation. Displayed are the scanned values for the ␣ 128/142 subunit bands (Fig. 4A). The data are plotted on a semilog scale. The half-life was estimated to be 10.9 h for ␣ 128/142 based on a least squares fit of an exponential function to the data, which is within the range of values found for the wild type subunits under the same conditions (8). placed by serines to avoid the possibility that aberrant disulfides form between either cysteine 192 or 193 and the remaining cystine loop cysteine. For the ␤ subunit mutation, ␤ 128 , the formation of the cystine loop was eliminated by replacing only cysteine 128 with serine. Each of the two mutated subunits, along with the other three wild type subunits, was stably transfected into L fibroblasts. Cell lines were isolated that expressed each of the mutated ␣ and ␤ subunits and the corresponding three wild type subunits (Figs. 4A, lane 1, and 5A,  lane 1). 125 I-BuTx binding experiments were performed on the stably transfected cells to characterize the effect of the mutations on AChR expression (Fig. 3, A and B). No 125 I-BuTx binding was detected on the cell surface of either the ␣ 128/142 ␤␥␦ or the ␣␤ 128 ␥␦ cell lines (Fig. 3A) or for the intracellular compartments of the ␣ 128/142 ␤␥␦ cells (Fig. 3B). However, intracellular 125 I-BuTx binding sites were expressed in the ␣␤ 128 ␥␦ cells (Fig. 3B). The intracellular 125 I-BuTx binding sites were immunoprecipitated using ␤ subunit-specific antibodies, which demonstrates that these sites in the ␣␤ 128 ␥␦ cells contained the mutated ␤ subunits.
Since the ␣ 128/142 ␤␥␦ cells failed to express any BuTx sites, we tested whether the mutated ␣ subunit assembled with other subunits. [ 35 S]methionine pulse-chase experiments were performed on the ␣ 128/142 ␤␥␦ cell line to characterize the assembly of the mutated subunit with the three wild type subunits. Displayed in Fig. 4A are the [ 35 S]methionine-labeled subunits from the ␣ 128/142 ␤␥␦ cells immunoprecipitated with ␣ subunitspecific polyclonal antibodies (lanes 2-6). As shown by the coprecipitation of the wild type ␤ and ␥ subunits with the ␣ 128/142 subunit, these two wild type subunits assemble with the mutated ␣ subunits in approximately a 1:1 ratio. No assembly was observed between ␦ subunits and the mutated ␣ subunits. These results indicate that ␣ 128/142 ␤␥ trimers assemble, but neither the ␦ nor second ␣ 128/142 subunits subsequently assemble with the ␣ 128/142 ␤␥ trimers. We were unable to precipitate any [ 35 S]methionine-labeled subunits with conformation dependent mAb 14 (data not shown); thus, the mAb 14 epitope does not form on the ␣ 128/142 ␤␥ complexes during assembly. Furthermore, BuTx binding sites failed to form on the ␣ 128/142 ␤␥ complexes (Fig. 3, A and B). Subunit assembly in the ␣ 128/142 ␤␥␦ cells is thus blocked at the same step as the block of assembly by DTT, i.e. after assembly of the ␣␤␥ trimers and before the BuTx binding site and mAb 14 epitope form.
Similar to the ␣␤␥ trimers assembled in the presence of DTT, the assembled ␣ 128/142 ␤␥ complexes degrade more rapidly than ␣␤␥ complexes in the absence of DTT. The faster rate of degradation appears to occur because the complexes are no longer stabilized by events subsequent to trimer formation (Fig. 2D). As shown in Figs. 4C and 5B, the rate of ␣ 128/142 and ␤ 128 subunit degradation is in the range found for the unassembled wild type subunits (8). Therefore, the failure to complete the assembly process is not caused by an increased rate of degradation of the mutated subunits.
In addition to the block of assembly, the efficiency of assembly of the ␣ 128/142 ␤␥ trimers was reduced 2-3-fold relative to  2-6), or mAb 14 (lane 7). As in Fig. 2, A and B, the band labeled as ␣Ј coprecipitates with the other subunits. The band (ϳ43 kDa) that migrates between the ␣ and ␤ subunit bands is believed to be actin. B, the rate of ␤ 128 subunit degradation. Displayed are the scanned values for the ␤ 128 subunit bands and the ␣ subunits that coprecipitated with the ␤ 128 subunits in Fig. 5A. The data are plotted on a semilog scale. The half-life was estimated to be 15.4 h for ␤ 128 subunits based on a least squares fit of an exponential function to the data, which is within the range of values found for the wild type subunits under the same conditions. C, sedimentation of ␣␤ 128 ␥␦ subunit complexes. ␣␤ 128 ␥␦ and ␣␤␥␦ cells were bound with 125 I-BuTx as in Fig. 3B and size-fractionated on a 5-20% linear sucrose gradient. Shown are 125 I counts from the 125 I-BuTx-bound intracellular complexes in fractions 4 -14, which were immunoprecipitated with the ␤ subunit-specific mAb 148. ␣␤␥␦ cell surface complexes were first bound with cold BuTx to block 125 I-BuTx binding to the surface AChRs. Also displayed on the figure are the standards, alkaline phosphatase (5.4 S), catalase (11 S), and surface ␣ 2 ␤␥␦ complexes (9 S; dashed line). The ␣␤ 128 ␥␦ complexes peak at 8 S as estimated from a least squares linear regression fit to the S values of the standards. The shape of the intracellular ␣␤␥␦ profile can be duplicated by the sum of the intracellular ␣␤ 128 ␥␦ profile reduced by 60% plus the 9 S peak cell-surface ␣ 2 ␤␥␦ complexes. This indicates that intracellular ␣␤ 128 ␥␦ complexes are sim-ilar in size and composition to intracellular ␣␤␥␦ complexes with the exception of the ␣ 2 ␤␥␦ complexes in the 9 S peak. The broad profile observed for the intracellular AChR complexes relative to the cell surface ␣ 2 ␤␥␦ 9 S peak has been seen in other studies both for the Torpedo subunits at reduced temperature (4,8,40) and the mouse subunits at 37°C (3,5,6,8). D, the effects of the ␣ 128/142 and ␤ 128 subunits on subunit assembly are consistent with the model shown. The ␣ 128/142 subunit and DTT block assembly after the formation of trimers but before the formation of the BuTx binding site and mAb 14 epitope. The ␤ 128 subunit blocks assembly after the addition of the ␦ subunit but before the addition of the second ␣ subunit. wild type subunit assembly (see Fig. 2A, lane 2, and Fig. 6C,  lane 1). A similar reduction in assembly efficiency was observed for subunit complexes containing the ␤ 128 subunit (Fig. 5A). The decrease in assembly efficiency could be caused by misfolding of some of the mutated subunits, which has been observed for other proteins where cysteines have been mutated (21). To examine whether ␣ 128/142 subunits are misfolded, we tested whether the ␣ 128/142 subunits are recognized by mAb 35. mAb 35 is a conformation-dependent antibody specific for the ␣ subunit (2). mAb 35 differs from mAb 14 in that its epitope forms on unassembled ␣ subunits, and it forms well before the mAb 14 epitope (8). As shown in Fig. 4B, the results obtained with the mAb 35 precipitation are similar to the results with ␣ subunit-specific polyclonal antibodies (Fig. 4A, lanes 2-6). mAb 35 recognizes a large percentage of unassembled ␣ 128/142 subunits as well as the ␣ 128/142 ␤␥ complexes, which indicates that the unassembled ␣ 128/142 subunits are not grossly misfolded.

Elimination of the Disulfide Bond on the ␤ Subunit Blocks Assembly at a Later
Step-The finding that the BuTx binding site forms on complexes that contain the ␤ 128 subunit (Fig. 3B) suggests that subunit assembly with the ␤ 128 subunit progresses to a later step than assembly with the ␣ 128/142 subunit. The assembly of the ␤ 128 subunit with the wild type ␣, ␥, and ␦ subunits was characterized using an [ 35 S]methionine pulse-chase protocol with the ␣␤ 128 ␥␦ cells as shown in Fig. 5A,  lanes 1-7. In contrast to subunit assembly with the ␣ subunit mutation or after DTT treatment, assembly continued after the formation of ␣␤␥ trimers. Immunoprecipitation of the labeled subunits with ␤ subunit-specific antibodies (lanes 2-6) coprecipitated the ␦ subunit as well as the ␣ and ␥ subunits. The mAb 14 epitope forms on the assembled complexes as shown by precipitation of the [ 35 S]methionine-labeled subunits by mAb 14 (Fig. 5A, lane 7), and as with the ␤ subunit-specific antibodies, all four subunits are precipitated. The BuTx binding site also forms on these complexes as shown by the 125 I-BuTxbound complexes precipitated by ␤ subunit-specific antibodies (Fig. 3B).
The assembly of ␣␤ 128 ␥␦ complexes differs from the assembly of the four wild type subunits in that the second ␣ subunit is not added to the ␣␤ 128 ␥␦ complexes. The pulse-chase experiments demonstrate that the amount of ␣ subunit coprecipitated with ␤ 128 subunits (Fig. 5B) is constant throughout the pulsechase protocol, in contrast to the doubling observed with wild type subunit assembly. To further investigate the nature of the ␣␤ 128 ␥␦ subunit complexes formed, their sedimentation on a linear sucrose gradient was determined (Fig. 5C). The ␣␤ 128 ␥␦ complexes, bound with 125 I-BuTx and immunoprecipitated with ␤ subunit-specific antibodies as in Fig. 3B, migrated in a peak at a value of 8.3 S. The size of the ␣␤ 128 ␥␦ complexes is consistent with tetramers (8), which sediment at ϳ8 S and are smaller than the cell-surface ␣ 2 ␤␥␦ complexes, which sediment at 9 S.
The effects of the two subunit mutations and DTT on subunit assembly are summarized in Table I and are consistent with the model displayed in Fig. 5D. Based on the results of both the pulse-chase experiment and the sedimentation of the mature ␣␤ 128 ␥␦ complexes, the end product of subunit assembly in the ␣␤ 128 ␥␦ cells is an ␣␤ 128 ␥␦ tetramer. Thus, the ␤ 128 subunit, like the ␣ 128/142 subunit, causes a block in subunit assembly. However, the ␤ 128 subunit blocks assembly at a later step, after the formation of ␣␤␥␦ tetramers and before the addition of the second ␣ subunit. The failure of ␣␤ 128 ␥␦ complexes to fully assemble into pentamers provides an explanation for why ␣␤ 128 ␥␦ complexes are not transported to the cell surface (Fig.  3A). Since ␣␤ 128 ␥␦ complexes never fully assemble, most likely they are retained and degraded in the endoplasmic reticulum (22).
The ␣ Subunit Cystine Loop Changes Conformation during Assembly-To address when the ␣ subunit cystine loop forms during assembly, we made use of a mAb (mAb 259) that selectively recognizes the ␣ subunit only when the cystine loop is intact (23). This specificity of the mAb is demonstrated in Fig.  6A, where the mAb failed to recognize either the reduced ␣ subunit or the mutated ␣ subunit with the cystine loop elimi-  3-8). In lanes 1 and 8, subunits were denatured with 1% SDS prior to the immunoprecipitation. The marked ␣Ј band, which migrates just above the ␦ subunit band, is different from the ␦ subunit. It is specifically precipitated by ␣ subunit-specific antibodies even after coprecipitating subunits are dissociated by the SDS treatment (lane 1) and is expressed in cells, the ␣␤␥ cells, which lack the ␦ subunit (lane 2). C, ␣␤␥␦ cells were pulse-labeled with [ 35 S]methionine and chased for the indicated times. Labeled subunits were precipitated with ␣ subunit-specific polyclonal antiserum. D, ␣␤␥␦ cells were pulse-labeled with [ 35 S]methionine and chased for the indicated times in the presence of 5 mM DTT. Labeled subunits were precipitated with ␣ subunit-specific polyclonal antiserum.
nated. In Fig. 6, B and C, the ␣␤␥␦ cells were pulse-labeled with [ 35 S]methionine and chased to test at what point in assembly the ␣ subunit is recognized by the cystine loop mAb. Immediately following the half-hour [ 35 S]methionine pulse, the cystine loop mAb precipitated about the same amount of labeled ␣ subunit as our ␣ subunit-specific polyclonal antibodies (Fig. 6B,  compare lanes 1 and 3). The ␣ subunit cystine loop, thus, must form shortly after the subunit is synthesized. Since the events blocked by DTT and by the ␣ subunit mutation occur several hours after the synthesis of the ␣ subunit, the block occurs after the formation of the cystine loop.
The ability of the cystine loop mAb to precipitate ␣ subunits diminished with time. The loss of the epitope occurs during subunit assembly as shown by the difference in the subunits precipitated by the cystine loop mAb compared with the subunits precipitated with the ␣ subunit-specific polyclonal antibodies (Fig. 6, B, lanes 3-7, and C, lanes 1-5). During the chase, the amount of ␣ subunit precipitated by the cystine loop mAb is progressively reduced. By the last chase time, the cystine loop mAb precipitated very little ␣ subunit (Fig. 6B, lane 7) and also was unable to precipitate any significant amount of cell surface AChRs (Fig. 7A). The mAb epitope, although inaccessible to the mAb, is still present on the ␣ subunit as demonstrated by the cystine loop mAb precipitating as much ␣ subunit as the ␣ subunit-specific antibodies after the subunits were denatured with SDS (Fig. 6, B and C, compare lanes 8 and 5). The loss of the epitope thus results from a conformational change that buries the epitope.
The Conformational Change Occurs before Formation of the BuTx Binding Site and the Addition of the ␦ Subunit-The only AChR subunit complexes that appear to be recognized by the cystine loop mAb are ␣␤␥ trimers. This is most clearly observed at the times when the ␦ subunit is maximally assembled with the other subunits, i.e. 24 -48 h after assembly begins (Fig. 2, A and C). At these times, only ␤ and ␥ subunits coprecipitate with the ␣ subunits recognized by the cystine loop mAb (Fig.  6B, lanes 6 and 7), although complexes containing ␦ as well as the other three subunits are present when all ␣ subunits are precipitated (Fig. 6C, lanes 4 and 5) or when the subunits are precipitated by the ␥-specific mAb or mAb 14 ( Fig. 2A). At earlier times in the pulse-chase experiments of Fig. 6, B and C, the ␣Ј band, which is specifically recognized by the cystine loop mAb and ␣ subunit-specific antibodies, may be obscuring the presence of any ␦ subunit. The DTT block of assembly was used to address whether ␦ subunits are present at the earlier times during assembly. As shown in Fig. 2B, the addition of 5 mM DTT blocks assembly so that only ␣␤␥ trimers assemble. In Fig.  6D, subunit complexes were precipitated with the ␣-specific antibodies after cells were treated with 5 mM DTT. There are no significant differences between the results in Fig. 6B (lanes  4 -7) and those of Fig. 6D. From this we conclude that the complexes recognized by the cystine loop mAb are identical to the ␣␤␥ trimers precipitated by the ␣-specific antibodies in Fig.  6D.
To further characterize the subunit complexes recognized by the cystine loop mAb, we tested whether the BuTx binding site and the mAb 14 epitope form on these complexes. The cystine loop mAb precipitated only about 12% of the intracellular 125 I-BuTx binding sites precipitated by either mAb 14 or the ␥ subunit-specific mAb alone (Fig. 7B). To ascertain whether these 125 I-BuTx binding sites are on complexes containing ␥ subunits and the mAb 14 epitope, 125 I-BuTx binding sites were first precipitated with either mAb 14 or the ␥ subunit-specific mAb before using the cystine loop mAb. The number of sites recognized by the cystine loop mAb was not significantly reduced if samples were first depleted of complexes containing ␥ subunits or the mAb 14 epitope (Fig. 7B). Therefore, the ␥ subunit-containing complexes recognized by the cystine loop mAb do not contain the BuTx binding site or the mAb 14 epitope. Since formation of the BuTx binding site and the mAb 14 epitope precedes assembly of ␣␤␥␦ tetramers (8), the com-  The point during subunit assembly when the cystine loop conformational change takes place is evident in the pulse-chase experiment of Fig. 6B. The loss of the cystine loop epitope during assembly, shown by decreasing amounts of ␣, ␤, and ␥ subunits that are precipitated by the cystine loop mAb, begins at the 3-h time point and is almost completed by the 24-h time point (Fig. 6B, lanes 4 -6). This is the time period during which the BuTx site and mAb 14 epitope form and are quickly followed by the addition of ␦ subunits to the trimers (Fig. 2, A and C; see also Ref. 8). The similarity in the kinetics of the cystine loop conformational change and these assembly events together with the block of these events by ␣ 128/142 and DTT demonstrate a strong correlation between the ␣ subunit cystine loop conformational change and the assembly events. As summarized in Table I, there is a close identity between the subunit complexes precipitated by the cystine loop mAb and those assembled in the presence of DTT and with the ␣ subunit mutation. Altogether the data indicate that after the ␣ subunit cystine loop forms, it undergoes a conformational change, which is essential for the BuTx site and mAb 14 epitope to form and for the ␦ subunit to associate with the ␣␤␥ trimer. DISCUSSION The strong conservation of the cystine loop among the different ionotropic neurotransmitter receptors and throughout evolution suggests that this structure plays a vital role with respect to the receptors. In this paper, we demonstrate that the cystine loop is essential for different events that take place during subunit assembly. Events occurring during the assembly of the AChR can be broadly classified as either subunit folding or oligomerization. Previously, we observed that AChR subunits continued to fold after associations with other subunits (8). The data suggested a link between subunit folding and subsequent subunit associations. We proposed that the folding events, BuTx binding site and mAb 14 epitope formation, are part of the process to create a recognition site for the incoming ␦ subunit. In support of this hypothesis, we find that eliminating the ␣ subunit cystine loop blocks these folding events and also blocks the subsequent addition of the ␦ subunit to the ␣␤␥ trimer. As shown in the model in Fig. 8, we suggest that a subunit recognition site for the ␦ subunit is created concurrently with the ␣ subunit cystine loop conformational change. Because elimination of the ␤ subunit cystine loop blocks the addition of the second ␣ subunit to the ␣␤␥␦ tetramer, we further propose that the recognition site for the second ␣ subunit is created in parallel with a change in the conformation of the ␤ subunit cystine loop.
Formation of the different subunit recognition sites is likely to involve large rearrangements of the assembly intermediates. Such a rearrangement occurs on the ␣ subunit. Three folding events, 1) the cystine loop conformational change, 2) the formation of the BuTx binding site, and 3) the formation of the mAb 14 epitope, occur at about the same time and precede the addition of ␦ subunits to ␣␤␥ trimers. The three events occur on three separate regions that span the length of the N-terminal domain of the ␣ subunit. The mAb 14 epitope overlaps or is very near to the main immunogenic region (24) at amino acids 67-76 (25,26), while the BuTx binding site is at the other end of this domain in the region of amino acids 185-196 (27, 28). Regions of the ␥ subunits are also involved in these events, since the presence of this subunit is necessary for the mAb 14 epitope to form and greatly enhances formation of the BuTx binding site (8). Since the ␣ subunit cystine loop conformational change is part of a large rearrangement of the ␣ subunit, it is possible that the cystine loop itself does not directly associate with the ␦ subunit during the assembly of the ␣␤␥␦ tetramer. Instead, other regions, involved in subsequent folding events and distant from the cystine loop, could associate with the ␦ subunit.
A feature of our model (Fig. 8) is that AChR subunit assembly is controlled by the ordered formation of subunit recognition sites. These events are rate-limiting, as shown by the kinetics of the ␦ and second ␣ subunit additions, and as such provide checkpoints during assembly where the fidelity of the assembly process can be tested. Each subunit recognition site would form only if assembly had proceeded properly up to that point. Any misfolded or misassembled intermediates would be prevented from participating in later stages of assembly, and these improperly assembled subunits would be selectively retained and targeted for degradation by the endoplasmic reticulum "quality control" mechanisms (29). This paradigm, where the formation of subunit recognition sites is rate-limiting and contributes to quality control mechanisms during oligomer assembly, is likely to be found in the assembly of other ionotropic neurotransmitter receptors and ion channels (1) and may also apply in the assembly of other complex oligomeric proteins. Key   FIG. 8. Cystine loop "switch" model. Based on our results, we propose the cystine loop "switch" model, where the formation of the subunit recognition sites for the ␦ subunit and second ␣ subunit depend on the conformation of the ␣ and ␤ subunit cystine loops as diagrammed. The subunit recognition sites are created by a series of folding events. The folding events that create the ␦ subunit recognition site include the formation of the BuTx site and the mAb 14 epitope and the change in conformation of the ␣ subunit cystine loop. The change in conformation of the ␣ subunit cystine loop is essential and occurs early in this chain of events. The cystine loop is modeled as a switch. In the cystine loop "up position," ␦ subunit recognition site formation and the rest of assembly are blocked. In the cystine loop "down position," subunit recognition site formation and assembly continue. A similar role for the ␤ subunit cystine loop is proposed for the formation of the second ␣ subunit recognition site.
to the formation of subunit recognition sites appears to be the cystine loop conformational change. As shown in Fig. 8, we envision the role of the cystine loop as a switch in subunit assembly. In the "up position," the cystine loop blocks subunit recognition site formation and assembly. In the "down position," subunit recognition site formation and assembly continue. In such a role, the cystine loop conformational change is an essential part of the assembly process, which allows the subunits themselves to guide proper subunit folding and maintains the correctly ordered pathway by which subunits oligomerize.