Incorporation of Acetylcholine Receptors into Liposomes VESICLE STRUCTURE AND ACETYLCHOLINE RECEPTOR FUNCTION*

Functionally intact acetylcholine receptors can be solubilized from electric organ membranes of Torpedo californica and incorporated into liposomes by the cholate dialysis technique. Freezing and thawing of the reconstituted preparation appears to seal a population of initially leaky vesicles and leads to vesicle fusion. Inclusion of supplementary cholesterol at an optimal concentration of 20% (w/w) greatly enhances vesicle fusion during the freeze-thaw cycle. Size analysis by electron microscopy of negatively stained preparations indicates that fusion is accompanied by shifts in size and volume distributions of the vesicle population. Liposomes formed in the absence of acetylcholine recep- tors are distributed over a substantially smaller size range than liposomes containing receptors. Acetylcho- line receptors appear in those liposomes as dimers of 80 A doughnut-shaped particles. Freeze-fracture repli- cas of reconstituted preparations reveal the presence of large vesicles containing particles which correspond in size to acetylcholine receptors and smaller liposomes devoid of particles. The distribution of particles in the reconstituted membranes is sparse compared to their dense packing in native electric organ membranes. The activation and desensitization of reconstituted acetylcholine receptors mediated by acetylcholine or carbamylcholine is dose dependent. The reconstituted receptors distinguish between these agonists in terms of binding affinity in a way similar to receptors in

Functionally intact acetylcholine receptors can be solubilized from electric organ membranes of Torpedo californica and incorporated into liposomes by the cholate dialysis technique. Freezing and thawing of the reconstituted preparation appears t o seal a population of initially leaky vesicles and leads t o vesicle fusion. Inclusion of supplementary cholesterol at an optimal concentration of 20% (w/w) greatly enhances vesicle fusion during the freeze-thaw cycle. Size analysis by electron microscopy of negatively stained preparations indicates that fusion is accompanied by shifts in size and volume distributions of the vesicle population. Liposomes formed in the absence of acetylcholine receptors are distributed over a substantially smaller size range than liposomes containing receptors. Acetylcholine receptors appear in those liposomes as dimers of 80 A doughnut-shaped particles. Freeze-fracture replicas of reconstituted preparations reveal the presence of large vesicles containing particles which correspond in size to acetylcholine receptors and smaller liposomes devoid of particles. The distribution of particles in the reconstituted membranes is sparse compared to their dense packing in native electric organ membranes.
The activation and desensitization of reconstituted acetylcholine receptors mediated by acetylcholine or carbamylcholine is dose dependent. The reconstituted receptors distinguish between these agonists in terms of binding affinity in a way similar to receptors in the native membrane. Correlation of the fractional occupancy of ligand binding sites by cobratoxin with inhibition of receptor function is used to demonstrate that in the reconstituted system the doubly liganded acetylcholine receptor prevails in controlling channel gating. The potential experimental advantages as well as limitations of this reconstituted system are discussed. Synaptic transmission at the vertebrate neuromuscular junction is mediated through the rapid release of acetylcholine from the nerve terminal, followed by its diffusion across a 50nm synaptic cleft and its binding to acetylcholine receptors located at the top of junctional folds of the postsynaptic membrane. Binding of acetylcholine to its receptor initiates a conformational change in this protein, which leads to the transient opening of large (conductivity -25 pS) relatively short lived (1-2 ms) cation selective channels, which depolarize the muscle membrane (for recent reviews, see Refs. 113). The acetylcholine-binding sites and the channel which they regulate are contained within the monomeric form of the AChR' (4,5 ) which is composed of 5 glycopeptides in the subunit stoichiometry cuZpy6 (6-8, reviewed

in Refs. 2 and 3).
Studies with monoclonal antibodies (9) and sequencing of the fiist 54 amino acids of all four subunits (8) have revealed considerable homology between the component polypeptides of the AChR. The observation that ligand binding and channel function can be preserved even after extensive proteolytic nicking of the AChR illustrates the intimate association between these subunits (10, ll). Affinity labeling with analogues of agonists (12)) or antagonists (13) and partial renaturation of the (Y subunit after subunit separation in sodium dodecyl sulfate (14) have indicated that the binding site for cholinergic ligands and neurotoxins is formed at least in part by the a subunit. Photoaffinity derivatives of local anesthetics, which act as noncompetitive blockers of the AChR channel were found to label the 6 subunit (15, 16). It is, however, still not known to what extent each of the subunits contributes to the structure of the ion channel or how agonist-mediated channel gating takes place at the molecular level.
Incorporation of purified AChRs into model membranes can provide systems in which functional consequences of structural alterations can be studied under controlled conditions. Several laboratories have c0ntribut.d to the reconstitution of functionally intact AChRs from "he electric organ of Torpedo californica into both lipid vesicles (17)(18)(19)(20)(21)(22)(23) and planar lipid bilayers (24-26; for a review see Ref. 27). In a recent publication, we described the reassembly of AChRs and soybean lipids into vesicles by dialysis of the anionic detergent sodium cholate (20). This study emphasized that the stability of the AChR channel during solubilization of electric organ membranes and during the reconstitution process is dependent on the presence of supplementary soybean lipids, a notion supported by similar reports from other laboratories (17, 28; see also Ref. 27). AChR reconstitution by cholate dialysis, when conducted under optimal lipid conditions, yields two products: relatively large AChR-containing vesicles of fairly uniform protein/lipid composition in which AChRs are about 5-fold less densely packed than in native electric organ membranes and smaller liposomes devoid of AChRs (20). In this paper we present a detailed description of the structural features of the reconstituted vesicles and the functional characteristics of the AChRs incorporated in these model membranes. As an extension to previous reconstitution protocols, cholesterol is introduced as an agent which enhances vesicle fusion during a freeze-thaw cycle of the reconstituted preparation. We further demonstrate that AChR function in our reconstituted system should be interpreted with caution in view of limitations imposed by the structure of the reconstituted vesicles and the assay system, which may generate distortions of actual AChR behavior. Finally, our results justify the conclusion that reconstituted AChRs function pharmacologically in a way similar to AChRs in native membranes.

MATERIALS AND METHODS
Preparation of Electric Organ Membranes-AChR-rich membranes from the electric organ of 2' . californica (Pacific Bio-Laboratories, Inc., Venice, CA) were prepared by sucrose gradient centrifugation following differential centrifugation of the initial homogenate after the method of Elliott et al. (29), as previously described (19). Peripheral membrane protein was removed by alkaline extraction, essentially as described by Neubig et al. (30). AChR concentration was measured by immune precipitation as the concentration of lZ51a-BGT-binding sites (31). Protein was determined according to the method of Lowry et al. (32) using bovine serum albumin as standard. The specific activity of the alkaline washed membranes was 3.2 f 0.2 pmol of IZ5I-aBGT binding sites per g of protein.
lncolporatwrz ofAChRs into Liposomes-Solubilization of electric organ membranes in mixed micelles of sodium cholate (2%, Interchem, MontluGon, France) and crude soybean lipids (5 mg/ml, L-a-phosphatidylcholine type IIS, Sigma) was conducted as described by Anholt et al. (20). After centrifugation of the extract at 165,000 X g for 30 min, the lipid concentration in the supernatant was adjusted to 25 mg/ml using a sonicated dispersion of soybean lipids (150 mg/ ml) in distilled water. Supplementary cholesterol could be included by mixing a 100 mg/ml solution of soybean lipids in chloroform with a 100 mg/ml solution of cholesterol (Sigma) in chloroform in the desired proportions. After evaporation of the chloroform under nitrogen or argon, the dry lipid film was suspended in a 2% sodium cholate solution in 100 mM NaCI, 10 mM Na phosphate buffer, pH 7.4, by sonication. Final concentrations of the components in the reconstitution mixture were 25 mg/ml of total lipid, 2% sodium cholate, 100 mM NaCl, 10 mM Na phosphate buffer, pH 7.4, and 1-2 PM AChR. Liposomes containing AChR were formed through removal of the detergent by dialysis for 16-18 h against 500 volumes of 100 mM NaCl, 10 mM NaN3, 10 mM Na phosphate buffer, pH 7.4, followed by dialysis for the same period against 500 volumes of 145 mM sucrose, 10 mM NaNa, 10 mM Na phosphate buffer, pH 7.4. The resulting reconstituted preparation was subjected to a freeze-thaw cycle by placing it in a freezer at -20 "C followed by thawing at ambient temperature.
Assay of AChR Function-AChR function in the reconstituted vesicles was measured at room temperature as the amplitude of the integrated uptake of "Na+ (New England Nuclear) during a 10-s incubation period in the presence of agonist. Background values were determined by substituting distilled water for the agonist solution in the assay. Entrapped "Na' was separated from external 22Na+ by passage through a Dowex 5OW-8X cation exchange column according to the method of Gasko et al. (33) as previously described (19). All assays were performed in triplicate. When acetylcholine was used as agonist, the assay was preceded by a 30-min preincubation period of the vesicles with 1 PM echothiophate iodide (Ayerst Laboratories, Inc., New York), an acetylcholine esterase inhibitor, which was generously donated by Dr. T. L. Rosenberry, Case Western Reserve University, Cleveland, OH. The apparent internal volume of the vesicles was measured by incubation of the reconstituted vesicles with "Na+ for 48 h at 4 "C prior to passage through Dowex. Measurement of AChR concentration and orientation in the reconstituted vesicles was conducted as previously described (19).
tions were diluted 20-fold in 145 mM sucrose, 10 m~ NaNu, 10 mM Na Electron XicroscGpy-For negative staining the vesicle prepmaphosphate buffer, pH 7.4, and a small volume was applied to a carbon filmed grid and stained with 2% aqueous uranyl acetate, pH 5.5. The samples were viewed at 80 kV on a JEOL lOOCX electron microscope equipped with a eucentric goniometer specimen stage and anti-contamination device.
For freeze-fracture the samples were fixed in 2% glutaraldehyde for 5 min, cryoprotected in 30% glycerol, and then rapidly frozen in Freon 22 cooled by liquid nitrogen. The samples were fractured, unidirectionally shadowed with platinum at an angle of 30°, and carbon stabilized using a Balzers 300 freeze-etch device equipped with a turbomolecular pump and quartz crystal thin film monitor. Stereo pairs of the freeze-fracture replicas were taken with the JEOL l00CX electron microscope using the eucentric goniometer stage at -5 and +5 degrees for subsequent three-dimensional examination.

RESULTS
Vesicle Structure-The investigation presented in this paper evolved from the initial observation that freezing and thawing of reconstituted vesicles, prepared as previousky described (20), results in a 2-to &fold amplification of the amplitude of the agonist-induced z2Na+ uptake response. Here we present evidence that this effect is due both to sealing of Isopycnic sucrose gradient centrifugation of reconstituted vesicles. Reconstituted vesicles were diluted with an equal volume of 1 0 0 mM NaCI, 10 mM NaN3, 10 mM Na phosphate buffer, pH 7.4, containing a trace label of lZ5I-aBGT (3.0 X lo-' M) and incubated for 1-2 h at 4 "C. Samples of 0.1 ml were applied to 4.8 ml of 5-25% (w/w) linear sucrose gradients and centrifuged to equilibrium in a Beckman SW 50.1 rotor at 200,000 X g for 18 h. Fractions of 165 pl were collected after puncturing the tubes and counted in a y-counter. The sucrose concentration of each fraction was determined at 20 "C with a refractometer. A, reconstituted vesicles before a freeze-thaw cycle. Notice the presence of a shoulder (arrow) on the otherwise symmetrical peak and the absence of extramembranous bottom of the gradient. B, the same reconstituted vesicles after a freeze-thaw cycle. Notice the virtual disappearance of the shoulder, a very slight shift in the position of the symmetrical peak toward the top of the gradient, and a very slight broadening of this peak. Liposomes, prepared in the absence of AChR, band between 10-12% (w/w) sucrose (20). a population of vesicles which were initially leaky, as well as to fusion of vesicles during the freeze-thaw cycle.
Previously we reported that isopycnic sucrose gradient centrifugation of reconstituted vesicles gave rise to a symmetrical peak banding at a characteristic density, indicating a fairly uniform protein/lipid composition of the reconstituted vesicles (20). Recently, when using AChRs derived from alkalineextracted membranes, such gradients revealed a previously nonresolved shoulder on the heavy side of this peak (Fig. 1A). A single cycle of freezing and thawing of the reconstituted preparation leads to the disappearance of this shoulder (Fig.   1B). We interpret this shoulder to represent a population of leaky membranes, which become tightly sealed upon freezing and thawing. This notion is supported by the observation that the apparent number of external 12'II-aBGT binding sites decreases as a result of the freeze-thaw cycle ( Fig. 2 A , inset). The amplification of the agonist-induced 22Na+ uptake response ( Fig. 2 A ) is paralleled by an increase in the apparent A, assay of "Na+ uptake, induced by M carbamylcholine. B, assay of apparent internal volume. The inset to A indicates the fraction of externally accessible '251-aBGT binding sites. 0, vesicles before freezing and thawing; 0, vesicles after freezing and thawing. The same samples were used for all three assays, and all measurements were made in triplicate. Error bars have been omitted where they are smaller than the symbol. The fraction of externally accessible Iz5I-aBGT binding sites before freezing and thawing was atypically high in these samples (for typical values see the legend to Fig. 12). The tendency of excess cholesterol to precipitate during cholate dialysis at cholesterol concentrations above 30% (w/w) renders measurements lo I A

UNBOUND
Iz5I-aBGT n 4 VOID VOLUME -I... internal volume of the vesicle preparation (Fig. 2B) and a shift in size distribution (Figs. 3-7, discussed below), suggesting that vesicle fusion also occurred during the freeze-thaw cvcle. appears to greatly enhance vesicle fusion during the freeze-membranes has no effect on the ""a' uptake response or on thaw cycle, an effect accompanied by an increase in the the apparent internal volume in the absence of a freeze-thaw amplitude of the agonist-induced "'Na' uptake ( Fig. 2, A and cycle (Fig. 2, A and 23). The enhancement of vesicle fusion Acetylcholine Receptor Reconstitution during a freeze-thaw cycle by cholesterol is maximal when the reconstituted vesicles are formed in the presence of 20% (w/ w) cholesterol. Inclusion of cholesterol during the incorporation of AChRs into liposomes tends to alleviate restraints on the preferentially right side out insertion of AChRs during the assembly of reconstituted vesicles, resulting in a more random orientation in the reconstituted membranes ( Fig. 2 A , inset). The parallel increase in amplitude of the "Na+ uptake response and in apparent internal volume suggests that the internal volume of the vesicles may to some extent limit the amplitude of the agonist-induced "Na+ uptake, a notion which will be elaborated on below.

-
Gel fitration of the reconstituted vesicles on Bio-Gel A-150m indicates the presence of a population of AChR containing liposomes of fairly uniform size (Fig. 3A). A freeze-thaw cycle of this preparation causes the appearance of an additional more heterogeneous population of vesicles of larger size, which elutes closer to and overlapping with the void volume ( Fig. 3B). A second freeze-thaw cycle does not further affect this elution profie and does not lead to an additional increase in the amplitude of the agonist-induced "Na+ uptake response (data not shown). Most of the vesicle population is recovered in the void volume when 20% (w/w) cholesterol is present as a component of the reconstituted vesicles during the freezethaw cycle (Fig. 3C). Electron micrographs of negatively stained reconstituted preparations are presented in Fig. 4. Before freezing and thawing, vesicles appear fairly uniform in size, but some vesicles appear to have burst leaving sheet-like remnants perhaps as a result of the staining procedure (Fig. 4A). After freezing and thawing, most vesicles appear intact and their size distribution appears somewhat more heterogeneous (Fig.  4C). The shoulder of the gradient in Fig. 1A may be associated with the population of fragile vesicles observed in Fig. 4A. Inclusion of cholesterol (20% w/w) in the reconstituted vesicles does not alter the appearance of the vesicle population before freezing and thawing (Fig. 4B) but during the freezethaw cycle cholesterol causes a striking increase in size heterogeneity leading to the appearance of large vesicles (Fig.   4D). Clumping of vesicles into groups is often observed. The tendency of vesicles to adhere together in solution would influence the elution patterns seen in Fig. 3, which should, therefore, be interpreted with caution. AChRs can be observed in the reconstituted membranes as characteristic doughnutshaped particles of approximately 80 A diameter, occurring primarily as dimers (Fig. 4, insets; see also Refs. 5 and 34). Not all AChRs may be visible in the electron micrographs in Fig. 4, since only AChRs oriented more or less perpendicular to the electron'beam will be discernible. Extramembranous AChRs are rarely seen. A size analysis of the reconstituted vesicles is displayed in Fig. 5. Liposomes formed in the absence of AChRs appear smaller than 600 A in diameter with a mean diameter of 340 A (Fig. 5A, upper panel). Freezing and thawing of this preparation shifts the mean diameter to 400 A (Fig. 5A, upper   panel). Cholesterol-supplemented liposomes without AChRs appear similar in size distribution to regular control liposomes (Fig. 5C, upper panel). However, a freeze-thaw cycle results in a significant shift of the mean diameter from 340 A to 600 A. About 8% of these liposomes reach a diameter larger than 1000 A after freezing and thawing (Fig. 5C, lower panel). In each case liposomes formed in the presence of AChR have larger mean diameters than the corresponding preparations formed in the absence of AChR (Fig. 5, B and C, closed arrows). Regular AChR-containing liposomes before freezing and thawing have a mean diameter of 520 A, and about 4% of the vesicles are larger than lo00 A in diameter (Fig. 5B, upper   panel). This mean diameter is averaged over both the population of AChR-containing liposomes and the population of pure lipid vesicles and, therefore, underestimates the average size of AChR-containing vesicles. Freezing and thawing of the preparation shifts the mean value of the diameter to 620 A and causes the appearance of a large number of vesicles greater than 1000 A in diameter (9%, Fig. 5B, lower panel).

C. CHOLESTEROL SUPPLEMENTED LIPOSOMES WITHOUT AChR
Cholesterol-supplemented liposomes containing AChRs initially have a size distribution similar to AChR-containing liposomes without supplementary cholesterol ( Fig. 5 0 , upper   panel), with an overall mean diameter of 520 A. After freezing and thawing, however, extensive fusion is evident from the shift in mean diameter from 520 8, to 760 A and the finding that 25% of the vesicles are now larger than 1000 A (Fig. 5 0 , lowerpanel). Increase in size heterogeneity is evident in each case in which fusion occurs and is most striking when supplementary cholesterol is present in the liposomes during the freeze-thaw cycle (Figs. 4 and 5). These fusion events appear more dramatic when the fraction of the total included volume is calculated for vesicles of a given diameter in each box of the histogram. In Fig, 6 we have plotted the histograms of Fig. 5,   B (both panels) and Lf (upper panel), as the percentage of the total included volume against vesicle diameter. In regular AChR-containing liposomes 50% of the total included volume is contained in about 20% of the total vesicle population with diameters larger than 760 A (Fig. 6 A ) . After a freeze-thaw cycle, 50% of the total included volume is contained in only -9% of the vesicle population having diameters larger than 1000 A (Fig. 6B). When supplementary cholesterol is present during the freeze-thaw cycle, 50% of the total included volume is contained in the few vesicles (3.5%) which are larger than 1400 A in diameter (Fig. 6 C ) . AChR-containing vesicles, before freezing and thawing, are expected to be of relatively large diameter (Fig. 5), and the extent to which fusion occurs during the freeze-thaw cycle is expected to be proportional to the area of the vesicle. Hence, the distribution of AChRcontaining vesicles can be predicted to shift toward large diameters and will experience relatively large volume increases (Fig. 6 ) . Fusion with liposomes can be expected to lower the packing density of AChRs in the reconstituted membranes, leading to a further increase in the available internal volume per AChR for ' " a ' accumulation during the agonist-induced permeability response (discussed in detail below).
Previously we reported that a monoclonal antibody which binds to a region on the extrasynaptic side of the AChR precipitates a fraction of the total '251-~BGT binding sites which closely corresponded with the percentage of externally accessible '"I-aBGT binding sites as determined by radioimmunoassay (20). We concluded from this experiment that  Fig. 5B, respectively. Histogram C corresponds to the lower panel of Fig. 5D. The histograms here were obtained by calculating the included volume in each box of the corresponding histogram of Fig. 5 and determining the fraction of the total integrated volume contained in the entire histogram which that volume represents. Arrows indicate the midpoints of the volume distributions.
AChRs in a single vesicle are oriented either all right side out or all inside out. Using the same assay we have observed that AChRs are oriented uniformly also in cholesterol-supplemented vesicles and that this uniform orientation of AChRs is preserved during a freeze-thaw cycle. This suggests that during the freeze-thaw cycle fusion events between AChRcontaining vesicles are relatively rare and that AChR-containing vesicles fuse primarily with the excess of AChR-free liposomes, leading to the formation of large vesicles containing uniformly'oriented AChRs at a lower packing density.
The contrast in packing density between AChRs in native and reconstituted membranes can be clearly seen in Fig. 7. Electron micrographs of freeze-fracture replicas of native membranes show numerous densely packed particles projecting from the plane of the membrane (Fig. 7A). In contrast, reconstituted vesicles contain far fewer sparsely distributed particles (Fig. 7, B and C). These particles are about 90 A in diameter and protrude at least 45 A from the plane of the membrane, as measured from the projected shadow of particles at the top of the vesicle. The dimensions of these particles correlate with the known structural dimensions of AChRs (34-37). In addition to large particle-containing vesicles small liposomes can be observed which do not contain particles (Fig.   7, B and C). These appear larger and more heterogeneous in size when the preparation has been frozen and thawed in the presence of supplementary cholesterol (Fig. 7C), as expected from the size distribution profiles shown in Fig. 5, A and C  (lower panels). Particles were never observed in liposome preparations prepared in the absence of AChR (data not Acetylcholine Receptor Reconstitution shown). Fig. 8 shows a stereoscopic negatively stained image of a native electric organ membrane, which clearly demonstrates the dense packing of AChRs in native membranes. Native membranes have sizes ranging between 2000 and 6000 8, in diameter, with a mean diameter of 4000 8, (histogram not shown). When viewed stereoscopically, the 4000 8, vesicle shown in Fig. 8, although collapsed, appears intact. Since the lipid matrix by this staining procedure appears transparent, AChRs can be seen in many orientations, both from the external surface and looking through the external surface of the vesicle at the inner surface of the opposite side of the membrane (arrows in Fig. 8). Rows of AChR dimers can be distinguished, and the majority of AChRs in this 4000 8, vesicle can be identified (a fraction of the AChRs (-15%) located on the sides of the vesicle is not visible). The vesicle in Fig. 8 contains approximately 1500 AChR monomers, which amounts to -1 AChR/335 nm' and -2 X loT2" liters (internal volume)/AChR. Previously we concluded from isopycnic sucrose gradient centrifugation that AChRs in reconstituted membranes are about 5-fold less densely packed than AChRs in native membranes (20). Hence, we would expect to find approximately 1 AChR/1700 nm' in a reconstituted membrane before freezing and thawing, i.e. -75 AChRs in a 2000 8, vesicle. Considering that AChRs make up about 7% by weight of the reconstituted membranes (20) and taking into account the partial specific volume of the AChR (6) AChR Function-It is generally agreed that AChRs can exist in a t least three different conformational states, a "resting state," in which the cation channel is closed, a "conductive state," in which the cation channel is open, and a "desensitized state," in which the cation channel remains closed even in the presence of bound agonist (1-3). Ligand binding to the AChR in its resting state results in opening of the channel, which remains open for several milliseconds, allowing the passage of cations. Prolonged exposure to the agonist will stabilize the AChR in its desensitized state, which is characterized by a higher affinity (by two orders of magnitude) for the ligand than the resting state (1-3, 39-44). Evidence suggests that the desensitization process occurs in two stages, a fast step (tlrP = 300 ms) which initiates the desensitization process, followed by a slow step ( t 1 / 2 = 6-7 s) which stabilizes the desensitized high affinity conformation (45-48; values for t l r 2 are cited from Reference 45 but will vary depending on agonist concentration, temperature, membrane voltage, and the concentration of divalent cations (reviewed in Refs. 1-3)). Recovery of the desensitized AChR to its resting state after removal of the agonist is a slow process occurring over minutes (1,40,(48)(49)(50). Since AChR activation and desensitization processes can go to completion within the incubation period of our assay (10 s), it is clear that our measurements of integrated '"Na' uptake are not sensitive to the intermediate kinetics which lead to the accumulation of the measured amount of '"a' inside the vesicles prior to the stabilization of a relatively long lived desensitized state. However, important qualitative information about the functional characteristics of reconstituted AChRs can still be obtained, for example through manipulation of the AChR concentration.
The dose response characteristics for AChR activation by agonists depend on the concentration of AChR (Fig. 9). When the AChR concentration is high with respect to the dissociation constant for the agonist, bound agonist will constitute a significant fraction of the total agonist concentration. The midpoint of the dose response curve will as a result be displaced as a function of AChR concentration and will asymptotically approach a constant value a t infinite dilution of

Acetylcholine Recel
AChR. This displacement effect is twice as large for acetylcholine as for carbamylcholine (Fig, 9), reflecting the higher affinity binding expected for acetylcholine than for carbamylcholine both to the resting and desensitized states of the AChR (40, 41). The effect of AChR concentration on the midpoint of the dose dependence of AChR desensitization is even more dramatic (Fig. lo), since the desensitized AChR has higher affinity (by two orders of magnitude) than the resting AChR for both agonists (40, 41). These observations justify the conclusion that AChR desensitization in the reconstituted system is also accompanied by a transition to a conformation of increased ligand binding affinity and that reconstituted AChRs display a greater affinity for acetylcholine than for carbamylcholine in both the resting and the desensitized state, as would be expected of these AChRs in the native membrane (40-43). Thus it appears from functional measurements that reconstituted AChRs distinguish between ligands and, at least qualitatively, undergo state transitions in a way similar to AChRs in the native membrane.
Correlation   (1y ) * (19, 51; see "Appendix"). In native membranes such functional measurements are obscured by the dense packing of AChRs (Fig. 8) and the concomitant limitation of internal volume/AChR (-2 X lo-'' liters/AChR), which leads to equilibration of the intravesicular volume with the external concentration of "Na+ before termination of the 22Na+ uptake assay. Hence, a large fraction of AChRs have to be occupied by cobratoxin before a reduction in the agonist-induced "Na+ uptake is observed, an effect previously described by us (19) and others (52) and illustrated in Fig. 11A. Previously we reported that in reconstituted vesicles which had not been passed through a freezethaw cycle titration of the "Na+ uptake response with cobra- toxin appeared linear (4, 19, 20). This led us to propose that control of channel gating is dominated by one of the two acetylcholine binding sites (191, in contrast with observations by other investigators (1, 51, 53, 54). Recent studies indicate, however, that the experimentally observed linearity is apparent, arising from a relatively small internal volume of the reconstituted vesicles limiting the agonist-induced "Na+ uptake response (-8 X lo-'' liters/AChR before a freeze-thaw cycle). A mathematical model of the cobratoxin titration experiment that takes into account different degrees of volume limitations on the "Na+ uptake response generates a family of curves which approach the parabola f ( y ) = (1y)' as restraints on the available internal volume per AChR are relieved ( Fig. 11B; for derivation see "Appendix"). It can be seen from Fig. 11B that at relatively low degrees of volume limitations per AChR (indicated by the parameter y as defined in the legend to Fig. 11B and in the "Appendix"), shallow sigmoids may within experimental error appear linear (see also Fig. 110). This apparent linearity is enhanced if one assumes heterogeneity in vesicle sizes, as observed in Fig. 5. Fig. 11, C-E, shows the range of observed toxin titration behaviors of regular reconstituted vesicles after freezing and thawing. The titration curves vary slightly, depending on the degree to which fusion has occurred as a result of the freezethaw cycle. Notice, however, that in Fig. 11, C and E , "Na+ uptake induced by a lower agonist concentration generates toxin titration curves shifted in the direction of the curve described by the equation, f ( y ) = (1y)*. Cholesterol-supplemented vesicles after freezing and thawing have relatively large internal volumes/AChR (-25 X liters/AChR) and show only slight deviations from this parabolic curve. Deviations are small at saturating concentrations of carbamylcholine, and the predicted titration curve is followed closely at 2 x M carbamylcholine, a concentration which elicits between one-third and one-half of the maximal response (Fig.  11F).
In native electric organ membranes about 75% of the total lZ5I-aBGT binding sites must be liganded by cobratoxin before inhibition of the "Na+ uptake response becomes evident (Fig.  11A). Under these conditions about 6% of the AChRs will have both of their agonist binding sites free of toxin. For a 4000 A native membrane vesicle, as in Fig. 8, this amounts to approximately 90 AChRs with -37 X lo-'' liters (internal volume)/AChR, ie. at M carbamylcholine equilibration is limiting in native membrane vesicles until the internal volume per active AChR is increased from -2 X lo-*' liters/ AChR to -37 X lo-'' liters/AChR. Reconstituted vesicles, supplemented with cholesterol after a freeze-thaw cycle contain -25 X lo-'' liters/AChR. The relative contributions of desensitization and volume limitations in determining the amplitude of the "Na' uptake response will vary with the concentration of the agonist. At low agonist concentrations when few AChRs are activated, desensitization is expected to be the primary limitation on the "Na' uptake. At higher agonist concentrations volume limitations will become important in an increasing fraction of the vesicles. In these vesicles equilibration of the intravesicular volume with the external "Na+ concentration will occur at submaximal agonist concentrations? Therefore, under these conditions we predict a shift in the midpoint of the dose response curve for AChR activation to lower agonist concentrations and a narrowing of the range over which saturation of the response occurs. The dose response curve of AChRs At saturating concentrations of carbamylcholine a third of the total intravesicular volume equilibrates with 22Na+ (Fig. 2B). However, since an unknown fraction of this volume is contained in AChRfree liposomes, more than a third of the volume of AChR containing vesicles is expected to equilibrate with "Na+ during the 22Na+-uptake response.
incorporated in cholesterol-supplemented liposomes broadens after a freeze-thaw cycle, and its midpoint shifts about 5-fold toward a higher agonist concentration, as would be expected from the removal of volume restrictions (Fig. 12).
We conclude that in reconstituted vesicles cobratoxin inhibition curves can be obtained which are consistent with those predicted by a model in which the doubly liganded AChR prevails in controlling channel gating. Limitations on internal volume per AChR may complicate investigation of AChR function but are much less severe in reconstituted vesicles (-8 t /  ( A ) or presence ( B ) of supplementary cholesterol, and dose response curves were obtained before (0) and after (0) passing the preparations through a freeze-thaw cycle. Carbamylcholine-induced uptake of "Na+ was assayed in tripicate for each point. Error bars are omitted where they are smaller than the symbol. Total AChR concentration was in each case 1.00 p~. In A the fraction of external '?-aBGT binding sites was 85% before freezing and 65% after a freezethaw cycle, whereas in B 73% of the AChRs were externally accessible before freezing and 50% after a freeze-thaw cycle.

DISCUSSION
Previously we have established optimal conditions for the stabilization of AChR channels in cholate solution and during the reconstitution process (20). Here we present an extension of these studies, in which we have characterized the structure of the reconstituted vesicles and the basic functional properties of the reconstituted AChRs. Our studies of vesicle structure are consistent with our previous observations (19,20) in that: 1) two populations of reconstituted vesicles are formed during cholate dialysis, one consisting of large liposomes containing AChRs and the other consisting of small liposomes devoid of AChRs (Fig. 7 , B and C), and 2) the packing density of AChRs in reconstituted vesicles is sparse compared to their packing density in native electric organ membranes (Fig. 7 ,  A-C).
In the course of our studies it became apparent that the agonist-induced 22Na+ uptake response is to some degree limited by a restricted intravesicular volume per AChR for the accumulation of 22Na+. The internal volume per AChR can be increased from -8 X lo-" liters/AChR to -25 X 10-" liters/AChR by subjecting the reconstituted preparation to a freeze-thaw cycle. Freezing and thawing appears to seal a population of previously leaky vesicles (Figs. 1 and 2, inset) and leads to vesicle fusion (Figs. 3 and 5). From Fig. 1, it appears that sealing of the reconstituted vesicles may account for approximately a 2-fold increase in the amplitude of the Na+ uptake response. The extent to which fusion occurs is variable, and we speculate that it may depend on the speed at which freezing takes place.
Inclusion of supplementary cholesterol during reconstitution greatly enhances subsequent vesicle fusion during freezing and thawing of the reconstituted preparation and randomizes the orientation of AChRs in the reconstituted membranes (Fig. 2). This observation suggests that previously reported effects of neutral lipids on AChR function (55, 56) may be accounted for by alterations in vesicle structure rather than modifications of AChR function per se. It is interesting to note that lipid extracts from AChR-enriched electric organ membranes contain 20% (w/w) cholesterol (57). Although we have not detected any direct influence of cholesterol on the function of reconstituted AChRs, we cannot exclude that effects may take place, such as shifts in the pre-existing isomerization equilibrium between desensitized and resting AChRs in the reconstituted membranes. At present no information is available regarding the pre-existing proportion of resting activatable AChRs compared to desensitized AChRs in the reconstituted membranes, a ratio which in native electric organ membranes ranges between reported values of 4.5

(40) and 10 (41).
Studies on AChR function in the reconstituted system under our assay conditions are complicated by the following factors: 1) restrictions on the available volume per AChR for the accumulation of "Na+ during the assay time; 2) AChR concentrations, which are frequently high relative to the dissociation constant for the agonist; and 3) measurement of an integrated "Na' uptake response rather than measurement of initial flux rates. Through the use of cholesterol-supplemented vesicles, volume restrictions can be minimized or, at low agonist concentrations, eliminated (Fig. 11F). The large response amplitude obtained after freezing and thawing of reconstituted vesicles allows measurement of AChR function at relatively low AChR concentrations, which have minimal effects on the midpoints of the dose response curves for AChR activation and desensitization (Figs. 9 and 10, closed circles). The major limitation on our current system remains the measurement of an integrated **Na+ uptake response which is insensitive to the intermediate kinetic steps involved in AChR function (e.g. during the assay time activation as well as desensitization processes take place, as for instance in Fig. 9). It is, for example, not possible with our present assay system to distinguish fast and slow desensitization processes (45)(46)(47)(48) or to make accurate correlations between agonist-mediated state transitions in terms of binding kinetics and functional behavior, as has been done in elegant experiments on AChRs in cultured BC3H1 muscle cells (39,58). Yet, despite these limitations we could obtain evidence which qualitatively supports the following conclusions. 1) Reconstituted AChRs distinguish correctly between agonists in terms of apparent binding affinity and undergo agonist-mediated conformational transitions between states of low and high affinity in a way similar to AChRs in the native membrane (39, 43). 2) The doubly liganded AChR prevails in the control of channel opening. T h e second conclusion is based on a careful analysis of cobratoxin inhibition curves and is in agreement with results obtained from similar experiments performed on AChRs in BCRHl cells (51). The cobratoxin titration experiment in our system is prone to distortions arising from internal volume limitations, which led us to conclude previously that liganding of a single agonist binding site dominates channel opening (19). Now, other investigators have also observed a parabolic decline of AChR function upon addition of increas- ing amounts of cobratoxin in reconstituted vesicles." Alterations in dose response characteristics as a result of volume limitations are significant but too small to affect our previous conclusions identifying the AChR monomer as functional unit in reconstituted vesicles (4) (dose response curves observed in this study happened in addition to be broad with midpoints at about 1.1-1.5 PM carbamylcholine, resembling the dose response curve for cholesterol-supplemented liposomes after freezing and thawing in Fig. 12B).
Limitations on time resolution imposed by our assay conditions can theoretically be alleviated through the use of rapid spectroscopic techniques or fast electrical measurements in planar bilayer systems. Rapid measurements through stopflow techniques have been performed by several investigators in order to study the kinetics of the cholinergic response in native membranes on the millisecond time scale (2, 3, 28,42-45, 47, 59). However, such techniques are still fraught with problems arising from internal volume limitations at saturating concentrations of agonist (44,47,59). An indirect spectroscopic method which relies on the fluorescence quenching of an entrapped dye as a result of agonist-induced influx of thallium ions, although elegant in design (59), suffered from a very low signal to noise ratio when applied to reconstituted vesicles (60). In contrast to stop-flow techniques, planar lipid bilayers allow direct electrical measurements of fast kinetic events in the millisecond time range and have the additional advantage that the opening and closing of individual AChR channels can be resolved. Incorporation of functionally active AChRs in planar lipid bilayers has been reported by several laboratories (24-26), and a preliminary report of the gating properties of single AChR channels has recently been presented (61). The mean channel open times were found to depend on the type of agonist used, but not on its concentration, and increased at hyperpolarizing voltages ( T = 4 ms at -70 mV in the presence of ~arbamylcholine).~ These studies indicate that AChRs in planar bilayers display electrophysiological characteristics which resemble those of AChRs in muscle membranes (1, 62). Since the AChRs incorporated in these planar bilayers are derived from reconstituted vesicles described in this paper, these observations strengthen our evidence that AChRs incorporated in liposomes are physiologically intact. In contrast to vesicles, however, planar lipid bilayers are less amenable to biochemical manipulation and are technically complex. Reconstituted vesicles, by virtue of their technical simplicity, are useful as a convenient and rapid screening system for macroscopic effects on AChR function. Combined studies of AChRs incorporated into liposomes and into planar lipid bilayers should provide a powerful approach for studying the molecular basis of AChR function.
In addition to the studies mentioned above, AChRs reconstituted into lipid vesicles have proven useful in other investigations. Using reconstituted vesicles two monoclonal antibodies were identified as noncompetitive inhibitors of AChRmediated cation permeability (63). Mapping the binding sites of such antibodies may help localize the cation channel and its linkage to the acetylcholine binding sites. Other experiments indicate great experimental potential for AChRs incorporated into Iiposomes as efficient immunogens in in vitro studies of "experimental autoimmune myasthenia gravis," an animal model of myasthenia gravis in which an autoimmune response against AChRs is elicited by immunization with purified AChR (64).5 Hence, reconstituted vesicles containing AChRs appear to be versatile and useful tools for various applications in AChR research.
Acknowledgments-We greatly appreciate the expert technical assistance of Derek Leong in preparing the freeze-fracture replicas.
We thank Peter Vasquez for excellent technical assistance.

APPENDIX
The following theoretical treatment which gives a qualitative description of the effect of limited internal volume on the cobratoxin titration curve has been simplified by the following assumptions. 1) AChR concentrations are assumed to be well below the dissociation constant of the agonist for the high affinity desensitized state. This allows us to ignore depletion of the total ligand concentration by AChRs that have cobratoxin bound to just one of their agonist binding sites. Such AChRs are nonfunctional with respect to "Na+ uptake but can still bind agonist removing it from the solution. 2) Vesicles are assumed to be of uniform size. 3) We assume that there is no voltage difference between the vesicle interior and the external solution. Since we know that the first two assumptions are violated in our experimental conditions, the simplified mathematical treatment given below should be considered only qualitative.
The rate of "Na+ uptake in reconstituted vesicles can be described by the equation duldt = n,P'(C,, -Ch) (1) in which u represents the number of moles of "Na+ inside the vesicles, n, the number of open channels, P' the effective channel permeability (liters/s),'j C, , the concentration of "Na+ in the external solution, and Cb the intravesicular concentration of "Na+ (M). In addition in which Vi, represents the intravesicular volume (liters). n, is  Based on our assumptions as listed above, n,(t) can be considered to be proportional to the number of AChRs that are functional (ie. capable of opening a channel). If we define the fractional occupancy of lz5I-aBGT binding sites by cobratoxin as y , such that y = the concentration of sites with bound cobratoxin/the total concentration of 1251-(rBGT binding sites ( y 5 l), then the fraction of AChRs with both ligand binding sites vacant will be equal to (1 -y)'. Hence, if both sites must bind agonist in order to open a channel, the number of open channels wiU be nJt) = (1y)'n,0(t) (4) in which n;(t) represents the number of open channels in the absence of cobratoxin. We can now define a parameter y, which describes the rate at which a given intravesicular velume f i s up with 22Na+ fluxing in through n; channels of f i e d permeability P', as follows.
' P ' is defined as the permeability P (cm/s) multiplied by the average cross-sectional area of a single channel (cm') x 10'. and normalizing u ( t ) with respect to the uptake of nNa+ in the absence of cobratoxin ( y = 0) will then give which is the function of y plotted in Fig. 11B. Observe that for y+ 0 (i.e. Vh + m), f ( y ) "* (1y) '. The graph of f ( y ) uersus y is convex for y 5 0.5 and sigmoidal for y > 0.5.