Quaternary associations of acetylcholinesterase. I. Oligomeric associations of T subunits with and without the amino-terminal domain of the collagen tail.

We investigated the production of acetylcholinesterase of type T (AChET) in COS cells during transient transfection. When expressed alone, Torpedo AChET remains essentially intracellular, forming dimers and tetramers; in contrast, rat AChET is secreted and produces mostly amphiphilic monomers (G1a) and dimers (G2a), together with smaller proportions of nonamphiphilic (G4na) tetramers, amphiphilic tetramers (G4a), and an unstable higher polymer (13.7 S). The latter two forms have not been described before. We show that secreted G1a and G2a forms differ from their cellular counterparts and that proteolytic cleavage occurs at the COOH terminus of “flagged” subunits. The binding proteins QN/HC and QN/stop are constructed by associating the NH2-terminal domain of the collagen tail (QN) with a functional or truncated signal for addition of a glycolipidic anchor (glycophosphatidylinositol). Coexpression with QN/stop recruits monomers and dimers to form soluble tetramers (G4na), increasing the yield of secreted rat AChE and allowing secretion of Torpedo AChE. Using antibodies against QN or addition of a flag epitope, we showed that the secreted tetramers contain the attachment domain. Coexpression with QN/HC modifies the distribution of AChET in subcellular compartments and allows the externalization of glycophosphatidylinositol-anchored tetramers at the cell surface.

We investigated the production of acetylcholinesterase of type T (AChE T ) in COS cells during transient transfection. When expressed alone, Torpedo AChE T remains essentially intracellular, forming dimers and tetramers; in contrast, rat AChE T is secreted and produces mostly amphiphilic monomers (G 1 a ) and dimers (G 2 a ), together with smaller proportions of nonamphiphilic (G 4 na ) tetramers, amphiphilic tetramers (G 4 a ), and an unstable higher polymer (13.7 S). The latter two forms have not been described before. We show that secreted G 1 a and G 2 a forms differ from their cellular counterparts and that proteolytic cleavage occurs at the COOH terminus of "flagged" subunits. The binding proteins Q N /H C and Q N / stop are constructed by associating the NH 2

-terminal domain of the collagen tail (Q N ) with a functional or truncated signal for addition of a glycolipidic anchor (glycophosphatidylinositol). Coexpression with Q N /stop recruits monomers and dimers to form soluble tetramers (G 4
na ), increasing the yield of secreted rat AChE and allowing secretion of Torpedo AChE. Using antibodies against Q N or addition of a flag epitope, we showed that the secreted tetramers contain the attachment domain. Coexpression with Q N /H C modifies the distribution of AChE T in subcellular compartments and allows the externalization of glycophosphatidylinositol-anchored tetramers at the cell surface.
The collagen-tailed forms of acetylcholinesterase (AChE, 1 EC 3.1.1.7) constitute a major component of the enzyme at neuromuscular junctions of higher vertebrates, mammals, and birds (for review, see Ref. 1). Their presence is controlled by innervation, by muscle activity, and by the type of muscle (slow or rapid) in a species-dependent manner. These molecules are hetero-oligomers, composed of AChE catalytic subunits of type T (AChE T ), associated with collagen subunits (Q). Each strand of the triple helical collagenic tail may be associated with a tetramer of T subunits. The AChE T subunits are linked by intersubunit disulfide bonds through a cysteine located at po-sition Ϫ4 of their COOH terminus: two "external" subunits are linked together, whereas the two "internal" ones are attached to the Q subunit (2)(3)(4).
We previously cloned the Q subunit from collagen-tailed AChE of Torpedo electric organs (5) and showed that it is able to associate with AChE T subunits of Torpedo, rat, or human AChE, forming collagen-tailed molecules in transfected COS cells (5)(6)(7)(8). The primary sequence deduced from the cDNA encoding Torpedo Q subunits comprises a signal peptide, an NH 2 -terminal domain (Q N ) which contains a pair of adjacent cysteines, a central collagen domain flanked by two pairs of cysteine residues, and a COOH-terminal domain (Q C ). Using antibodies directed against Q C , we showed that this COOHterminal region could be removed by collagenase without disrupting the assembly of catalytic tetramers. This experiment suggested that AChE T subunits were linked to the Q N domain (8). We further showed that an isolated Q N domain was sufficient to bind one AChE T tetramer, by constructing a chimeric protein in which Q N is fused to the COOH-terminal glycolipid (GPI) addition signal of the H subunit of Torpedo AChE (H C ). Coexpression of this Q N /H C protein with catalytic T subunits of Torpedo, rat, or human AChE produces GPI-anchored AChE tetramers (8,9).
We analyzed the enzyme produced by transfected COS cells expressing the rat AChE T subunit alone, exploring the effect of various parameters on activity and molecular forms. In agreement with a previous study (6), transfected COS cells expressing the rat AChE T subunit produce and secrete amphiphilic monomers and dimers of type II, G 1 a and G 2 a , (10 -12) and nonamphiphilic tetramers, G 4 na . We now show that the proportions of the various molecular forms may vary with culture, extraction, or storage conditions. We also show that, in addition to previously characterized molecules, the cells produce an amphiphilic tetramer, G 4 a , as well as an unstable component of 13.7 S.
We then investigated the effect of cotransfection with vectors encoding the Q N /H C protein, in which the attachment domain from AChE-associated collagenic subunits is fused to a GPI addition signal, or the Q N /stop protein, which does not possess this signal. Heteromeric associations of AChE T with these binding proteins produced tetramers that were attached to the plasma membrane by a GPI anchor (GPI-G 4 ) in the first case and secreted into the medium (G 4 na ) in the second case. We therefore analyzed AChE in both cell extracts and culture media using sedimentation, nondenaturing electrophoresis, and immunofluorescence of the transfected COS cells. Although most experiments were performed with rat AChE T , we obtained similar results with Torpedo AChE T . We found that the presence of Q N induces AChE T monomers and dimers to form tetramers, with which the attachment protein remains associated, even after cleavage of the GPI anchor.

EXPERIMENTAL PROCEDURES
Materials-All reagents were purchased from Prolabo (Paris, France) or from Sigma (St. Louis, MO, U. S. A.). PI-PLC from Bacillus thuringiensis was from Immunotech (Marseille, France). The M1 and M2 monoclonal antibodies directed against the "flag" epitope were from Eastman Kodak Co.
Expression Vectors and Site-directed Mutagenesis-The pEF-BOS expression vectors containing the coding sequence of the Torpedo Q subunit or of the chimeric Q N /H C protein were described previously (9). In the Q N /H C protein, the sequence encoding the Q N domain (numbered 1-110) was attached by a linker containing a BamHI site (encoding the residues GI, not numbered) to the sequence encoding the COOH-terminal peptide of Torpedo AChE H , H C (numbered 532-566); the truncated Q N /stop protein was obtained by site-directed mutagenesis, inserting a TGA stop codon with the single strand method (13), at position 551 within the sequence encoding H C . The cDNAs encoding Torpedo and rat AChE T were also inserted in pEF-BOS. Insertion of a sequence encoding the flag peptidic epitope was performed by mutagenesis in Q N /H C and in rat AChE T . In the case of Q N /stop, the peptidic epitope was added at the end of the Q N domain, after the linker G residue.
Transfection of COS Cells-COS cells were transfected by the DEAEdextran method, as reported (5), using 5 g of DNA encoding the catalytic subunit AChE T and various amounts of DNA encoding the binding protein, as specified. In the case of Torpedo AChE, the cells were allowed to recover for 2 days at 37°C after transfection and then transferred to 27°C, to allow production of active enzyme, for 2-3 days. In the case of rat AChE, the cells were maintained at 37°C and extracted 2-4 days after transfection. The culture medium (7 ml/10-cm dish containing about 5 ϫ 10 6 cells) was collected after variable periods of time, as indicated, for analysis of released AChE activity. The extracts and culture media were stored at Ϫ80°C.
Electrophoresis in nondenaturing polyacrylamide gels was performed as described previously (11). The gels contained 0.25% Triton X-100 and 0.05% deoxycholate and were electrophoresed for about 2-3 h under 15 volts/cm, with refrigeration at 15°C. AChE activity was revealed by the histochemical method of Karnovsky and Roots (15).
Production of Anti-Q N Antiserum-The rabbit polyclonal antiserum directed against the Q N domain was prepared against a recombinant protein produced in E. coli, as described (9).
Treatment with PI-PLC-Samples of extracts (50 l) were treated with 0.025 IU of PI-PLC for 1 h at 30°C.
Immunofluorescence-Immunofluorescence was performed with a rabbit antiserum directed against rat AChE (16), as described previously (17), except that the second antibody was an fluorescein isothiocyanate-conjugated F(abЈ) 2 fragment of anti-rabbit IgG (Silenius, Australia). The cells were analyzed either intact or after permeabilization with 0.2% saponin.

Quaternary Structures of Rat AChE T Subunits Produced in COS Cells
Expression of the Rat AChE T Subunit in Transfected COS Cells: Variable Patterns of AChE Molecular Forms-We studied the production of AChE in COS cells expressing the rat AChE T subunit. The activity obtained per culture dish increased with the cell density but plateaued at a maximal value around 10 6 cells/10-cm dish, and we therefore used this density in all of our experiments. Following transfection, AChE activity was barely detectable in the cells after overnight incubation and then increased steadily for at least 5 days. It appeared slightly later in the culture medium, where its proportion increased from about 30% of the total activity after 30 h to 70% after 4 days. In addition, the yield of cellular activity showed a saturable Michaelis-type increase with the amount of recombinant plasmidic DNA used for transfection, reaching a plateau at about 10 -15 g/dish, with half-saturation around 5 g of DNA/dish (not shown).
We analyzed the molecular forms of AChE in the cell extracts and in the culture medium, by sedimentation in sucrose gradi-  (Fig. 1, A and B). The sedimentation profiles did not change with the cell density or with the amount of DNA encoding AChE T /dish (not shown). In Fig. 1A, typical sedimentation profiles of cell extracts show the presence of monomers (G 1 a ; 3 S), dimers (G 2 a ; 4.5 S), amphiphilic tetramers (G 4 a ; 8.5 S), nonamphiphilic tetramers (G 4 na ; 10.5 S), and a 13.7 S molecular species. The proportion of oligomers to monomers increased progressively during the first days following transfection, and the sedimentation pattern then remained essentially constant between 3 and 5 days. The sedimentation profiles were reproducible in a given series of experiments, but we observed a rather large variability in the proportions of molecular forms (i.e. in the relative proportion of monomers to dimers and higher oligomers) when comparing experiments performed over a long period of time, possibly because of variation in the batches of cells. The cause of this variability was not investigated systematically. It was less marked in the medium than in cell extracts (Fig. 1B); the medium generally contained about 35% G 1 a , 45% G 2 a , 20% G 4 na , together with small variable proportions of G 4 a and sometimes of the 13.7 S form. In some cases we observed the presence of a nonamphiphilic monomeric form (G 1 na ), probably generated by a proteolytic process, after prolonged culture, e.g. 4 days (see below). Fig. 1 illustrates the influence of detergents on the sedimentation of the different molecular forms: amphiphilic forms sediment faster in the presence of Triton X-100 than in the presence of Brij-96, whereas nonamphiphilic forms are unaffected. Both monomers (G 1 a ) and dimers (G 2 a ) are amphiphilic forms of type II (10 -12), and the G 4 a form probably also belongs to this category, as evidenced by the influence of detergents on their electrophoretic migration (not shown) and on their sedimentation in the presence of Triton X-100 and Brij-96, respectively: 4.5 and 3 S (G 1 a ), 6.5 and 4.5 S (G 2 a ), 10 and 8.5 S (G 4 a ). Sedimentation profiles obtained in the presence of Triton X-100 or Brij-96 or without detergent showed that the 13.7 S component is not amphiphilic (Fig. 1A). Assuming that it is a globular protein composed exclusively of AChE subunits, the ratio of its mass to that of G 4 na tetramers, which is close to 1.5 ((13.7/10.5) 3/2 ), suggests that it is hexameric. This component is relatively abundant in fresh cellular extracts that had been maintained in the cold, without detergent (Fig. 2). In the absence of detergent it dissociated at 37°C, mostly into monomers (G 1 a ). In the presence of Triton X-100, dissociation occurred even in the cold and produced amphiphilic tetramers (G 4 a ) as well as monomers (G 1 a ). This instability explains why the proportion of the 13.7 S peak was variable and why this component did not generally appear in the culture medium. We also observed some inactivation of the G 1 a form at 37°C, particularly in the presence of Triton X-100; the instability of this form probably contributes to the variability of its proportion in cell extracts, especially when analyzed by nondenaturing gel electrophoresis.
Distinction between Cellular and Secreted AChE Molecules-Although the cellular and released forms cannot be distinguished readily by their sedimentation, they migrate slightly differently in nondenaturing electrophoresis (Fig. 3), possibly because of differences in glycosylation, or proteolytic cleavage, or other post-translational modification. To investigate the last possibility, we introduced by mutagenesis the peptidic sequence DYKDDDDK (flag epitope) after the normal COOH terminus of the rat AChE T subunit. The presence of this peptide at the COOH terminus did not significantly modify the production of active AChE, its distribution in the cellular and secreted fractions, or the proportions of its molecular forms. We found that although the cellular G 1 a and G 2 a forms carried the epitope, as shown by the effect of an anti-flag M2 monoclonal antibody on their sedimentation and electrophoretic migration, their secreted counterparts did not (not shown). This shows that proteolytic cleavage removes the flag peptide but not necessarily that it occurs within the COOH-terminal T peptide.
In fact, polyclonal antibodies raised against the last 10 residues of the T peptide of Torpedo AChE 2 or of rat AChE (18) were able to retard the electrophoretic migration of secreted as well as cellular active molecules (G 1 a , G 2 a , G 4 a , and G 4 na ), in the case of both species (not shown).

Coexpression of Rat and Torpedo AChE T Subunits with the Binding Proteins Q N /stop and Q N /H C
The structure of the binding proteins Q N /H C and Q N /stop is shown in Fig. 4. In cotransfection experiments, we used a nonsaturating dose of DNA encoding the rat AChE T subunit (5 g/dish) together with various doses of DNA encoding the binding proteins Q N /H C or Q N /stop (Q N /stop551).
Coexpression with Q N /H C Carries Rat AChE T to the Cell Surface-Immunofluorescence of transfected COS cells, with or without permeabilization, showed that AChE T was not associated with the cell surface when expressed alone. In contrast, a Q N /H C protein that included a flag peptidic epitope at its NH 2terminal extremity (see below) and could thus be visualized with a specific monoclonal antibody (M1), was exposed at the cell surface (not shown). When AChE T was expressed together with Q N /H C , it was carried with it to the cell membrane (Fig. 5) from where it could be released by PI-PLC (not shown), as in the case of GPI-anchored dimers generated from AChE H subunits 3 (19). The distribution of intracellular AChE was quite different in the two cases, showing an accumulation in vesicles in the presence of Q N /H C and a reticular pattern in its absence. This suggests that AChE T transits differently from the endoplasmic reticulum to the external medium when associated with a GPI-anchored protein. Fig. 6 illustrates sedimentation profiles of AChE produced at different times in cells expressing rat AChE T and Q N /H C ; these patterns show that tetramers accumulate progressively after transfection, between 1 and 2 days, whereas dimers and monomers increase little during the same period. This is entirely consistent with the fact that the metabolic half-life of monomers and dimers is much shorter than that of higher oligomers (20). To study the assembly of heteromeric tetramers with Q N /stop or Q N /H C , we analyzed the molecular forms produced after a delay of 2-4 days following transfection.

Conditions for Coexpression of Rat AChE T and the Binding Proteins Q N /stop and Q N /H C : Influence of the Levels of Expression Vectors-
We studied the effect of varying the amount of DNA encoding the Q N /stop or Q N /H C binding proteins, while the amount of DNA encoding AChE T remained constant (Fig. 7, A and B. In both cases, the cellular activity was increased at low doses of DNA encoding the binding proteins, but the total AChE activity produced per dish, in the cells and in the medium, was considerably reduced at high doses. At 40 g of DNA/dish, it was reduced to less than 10% of the control activity obtained when cells were only transfected with 5 g of DNA encoding AChE T / dish (not shown). In the case of Q N /stop, the proportion of secreted activity was increased significantly at all concentrations of DNA (about 95% of the total activity compared with about 70% in control cells expressing AChE T alone), in agreement with the production of soluble heterotetramers, and this was correlated with a much higher total activity/dish (Fig. 7B). In the case of Q N /H C , which forms GPI-anchored tetramers attached to cell membrane, the proportion of secreted activity decreased, at least up to 1 g of Q N /H C DNA (Fig. 7A). The activity recovered in the culture medium increased, however, at higher doses of DNA, reflecting the release of nonamphiphilic monomers (G 1 na ), as discussed below. Sedimentation profiles of corresponding cell extracts and culture media are shown in Fig. 8. The activity of G 1 a , G 2 a , and 13.7 S decreased in the cells and in the medium with the dose of DNA encoding either Q N /H C or Q N /stop. With increasing doses of Q N /stop, the activity of heteromeric tetramers, G 4 na , increased in the cells and in the medium with a concomitant decrease of G 1 a and G 2 a , reaching a maximum around 2-3 g of DNA encoding the binding protein/dish (Fig. 8, C and D); the decrease observed at higher doses reflected the decrease in total cellular AChE activity. In the case of Q N /H C , we observed a similar evolution of GPI-G 4 a in the cells (Fig. 8A). This GPI-G 4 a form was accompanied by a smaller proportion of G 4 na , probably representing a lytic derivative from which the GPI anchor was removed and which was also released in the medium (Fig. 8B). At high doses of DNA encoding Q N /H C , we observed the appearance of a nonamphiphilic monomeric form (G 1 na ), both in cell extracts and culture media, which cosedimented with G 2 a in the presence of Brij-96. This form was abundant in the experiment illustrated in Fig. 8, A and B, but its proportion was variable, and it was not always observed in other batches of COS cells. It probably results from a lytic process in the metabolic pathway of the GPI-anchored molecules, since it did not appear in the case of parallel cotransfections with Q N /stop, even at high doses of DNA (Fig. 8, C and D). Fig. 7 also illustrates the variation of the ratio R of heteromeric forms to total cellular or secreted AChE activity: R varies with the dose of DNA encoding Q N /stop (d), as R ϭ R(0) ϩ R max d/ (E ϩ d), where R(0) is the ratio obtained with AChE T alone, and R max is the maximal ratio obtained with a saturating dose of DNA. The parameter E thus defines an overall "efficiency" of interaction between the AChE T subunits and the binding proteins in the secretory pathway. Under specific experimental conditions, this parameter allows an evaluation of such interactions in living cells. In the case of Q N /stop (Fig. 7B), 3 S. Bon, F. Coussen, and J. Massoulié, manuscript in preparation. E is smaller for secreted than for cellular AChE, in agreement with the fact that the G 4 na form is exported more efficiently than the G 1 a and G 2 a forms, whereas in the case of Q N /H C , the values of E were approximately equal for the cellular and secreted compartments (Fig. 7A).
In the following experiments we generally used 5 g of DNA encoding AChE T and 6 g of DNA encoding the binding protein under study, unless indicated otherwise, as a compromise between the yield of total AChE activity and the proportion of hetero-oligomers obtained by association of the two types of subunits.
Do Released Tetramers Contain the Q N Domain?-Although GPI-anchored tetramers necessarily incorporated a processed (glypiated) Q N /H C protein, this was not obvious for the corresponding released G 4 na molecules obtained in the presence of Q N /H C or of Q N /stop. To detect the presence of the Q N domain, we used a rabbit polyclonal antibody raised against a fusion protein containing this domain. This antibody, anti-Q N , increased the sedimentation coefficient of both cell-associated and released tetramers produced by cells expressing Torpedo AChE T together with Q N /H C by about 1 S unit, whereas the G 2 peak was not affected (not shown). The effect of the antiserum was more marked in nondenaturing electrophoresis, probably because dilution was less important in this case (Fig. 9): a fraction of the G 4 form, secreted by cells coexpressing rat AChE T and either Q N /H C or Q N /stop, was retarded, indicating that the released tetramers contained the Q N binding domain.
To examine whether the complete binding protein remained associated in the hetero-oligomers, we also introduced the flag peptide at the NH 2 -terminal and at the COOH-terminal extremities of the Q N domain.
In the first case, the epitope was placed between residues Ala-42 and Glu-43 of the precursor sequence in the Q N /H C chimeric protein ("N-flagged" Q N /H C ), corresponding to the most likely cleavage site of the signal peptide (21). By coexpression of N-flagged Q N /H C with rat AChE T , we obtained cellassociated GPI-anchored G 4 a and released G 4 na , with the same apparent affinity as with Q N /H C , demonstrating that the presence of the flag epitope did not interfere with the heteromeric association of Q N and AChE T subunits. The sedimentation of these molecules was not affected visibly by the M1 monoclonal antibody, probably because its affinity was not sufficient to withstand dilution in the gradients, but they were both partially retarded by M1 during electrophoretic migration in nondenaturing polyacrylamide gels (Fig. 10A). Only a fraction of the molecules was retarded, as also observed in the case of anti-Q N , perhaps because accessibility of the epitope was restricted by the presence of an associated tetramer of catalytic subunits. Alternatively, it is possible that several sites of cleavage coexist or that the flag sequence introduced a new cleavage site, which eliminated the epitope from the mature protein, so that only a fraction of hetero-oligomers possessed a complete epitope.
In any case, this showed that the Q N domain was included not only in the GPI-anchored G 4 a molecules but also in the na . In addition, it indicated that the cleavage of the signal peptide occurred after Ala-42 at least in a fraction of the protein, since the M1 antibody is considered to recognize the flag epitope only in an NH 2 -terminal position. Glu-43 is therefore the likely NH 2 -terminal extremity of the mature Q N domain that binds an AChE T tetramer.
The flag epitope was also added immediately after the Q N domain, instead of H C . When this construction (Q N /flag) was cotransfected with rat AChE T , we obtained the same result as with Q N /stop, i.e. a large production and release of G 4 na , at the expense of G 1 a and G 2 a . In the medium the released G 4 na molecules contained the Q N /flag protein, as shown by the M2 antibody, which recognizes the epitope in a COOH-terminal position; it induced a shift in sedimentation (not shown) and retarded migration in nondenaturing electrophoresis (Fig.  10B). The presence of two distinct bands that react with the antibody suggests that at high doses of DNA, Q N may also be associated with dimers of AChE T subunits, in addition to tetramers.
Thus, the presence of a flag epitope, either in an NH 2 -terminal or a COOH-terminal position, did not prevent the production of the heteromeric AChE tetramers. The M1 antibody was able to recognize the NH 2 -terminal flag in the lytic cellular and released soluble tetramers (G 4 na ) derived from GPI-anchored G 4 a . Soluble tetramers obtained in the presence of Q N /stop carrying an NH 2 -or COOH-terminal flag were recognized by M1 and M2, respectively, showing that they also contained the binding protein.

DISCUSSION
The Fate of AChE T Subunits Produced in COS Cells: Quaternary Associations; Secretion-In the first part of this study we analyzed the production of AChE by transfected COS cells expressing rat AChE T subunits alone. As shown previously, the cells contained monomers (G 1  a ), dimers G 2 a , and tetramers (G 4 ) (6). The G 1 a and G 2 a forms correspond to amphiphilic molecules of class II, which interact with detergent and lipid micelles, probably through an amphiphilic ␣-helical region in the COOH-terminal T peptide 2 (1). The production of AChE increased with the quantity of DNA used for transfection, in a saturable manner, but the proportions of the different forms were not modified. AChE activity increased with time after transfection, with an accumulation of tetramers over monomers and dimers, in agreement with the fact that the metabolic turnover rate of the latter is more rapid (20).
We show here the existence of amphiphilic tetramers (G 4 a ) as well as nonamphiphilic tetramers (G 4 na ). As shown in the case of butyrylcholinesterase (22,23), the G 4 na component probably represents homotetramers of AChE T subunits, in which the amphiphilic helices of the T peptides interact with each other. In contrast, the structure of G 4 a might represent another type of quaternary organization in which all or part of the amphiphilic helices is exposed. The fact that this molecular form can be solubilized readily in the absence of detergent indicates that it differs from the membrane-bound G 4 a AChE of mammalian brain, which contains a 20-kDa hydrophobic anchor (24 -26).
COS cells also produced a nonamphiphilic, unstable component sedimenting at 13.7 S, which may be an hexamer of catalytic subunits, unless it contains other subunits. In spite of its instability, the 13.7 S form was sometimes observed in the culture medium, although in a minor proportion, suggesting that it does not incorporate intracellular resident proteins. It was also found in Xenopus oocytes expressing the rat AChE T subunit. 4 This component readily dissociates at 37°C, producing mostly monomers (G 1 a ) in the absence of detergent, but also dimers and amphiphilic tetramers (G 4 a ) in the presence of Triton X-100. Interaction with detergent micelles may stabilize a quaternary conformation of tetramers in which hydrophobic surfaces are exposed.
Finally, we illustrate the fact that, although the characteristics of the molecular forms are well defined, their proportions 4 E. Krejci, personal communication.
In the case of Q N /H C , the variation of R in the medium only tends to plateau if we include the lytic nonamphiphilic G 1 na form (R ϭ (G 4 a ϩ G 4 na ϩ G 1 na )/(total secreted activity), clearly suggesting that G 1 na was derived from heteromeric GPI-anchored tetramers. The values of E used for fitting the data were 0.7 for both cells and medium in the case of Q N /H C , 1 for cells and 0.15 for medium in the case of Q N /stop. are variable among different experiments, especially over long periods of time, using different batches of COS cells. Such variations may result from differences in the biosynthetic capacity of the cells, in the culture medium, or in the extraction and storage of the enzyme, especially in view of the instability of some of the molecular forms. The fact that the pattern of molecular forms may be modified in a temperature-dependent manner by detergent (Triton X-100) indicates that it does not exactly reflect the state of quaternary associations of AChE in the intact cell.
COS cells expressing Torpedo AChE T subunits produce a very small proportion of monomers, but mostly dimers, G 2 a , together with minor G 4 a and G 4 na forms. This suggests that, unlike rat AChE T , the Torpedo AChE T subunits are unstable in the monomeric state. The release of Torpedo and rat AChE in the culture medium differed markedly. The culture medium of COS cells expressing rat AChE T subunits contained about 30% of the total AChE activity after 2 days and 60% after 3 days following transfection, with similar proportions of amphiphilic dimers (G 2 a ) and monomers (G 1 a ) of type II, together with a smaller proportion of nonamphiphilic tetramers (G 4 na ). Although their sedimentation properties appeared identical, the migration of the cellular and released G 1 a and G 2 a forms was clearly different in nondenaturing electrophoresis. This suggests that post-translational modifications accompany the release of rat G 1 a and G 2 a AChE, which The amount of DNA encoding the binding protein, per dish, was the following: E, none; q, 0.1 g; Ç, 1.5 g; å, 15 g (see inset). In each case, the profiles are represented on the same scale so that the activities of the molecular forms obtained at the different doses of vector DNA may be compared directly. Note that the scale of activity is the same in panels A-C but different in panel D because secretion of G 4 na is increased considerably in the presence of Q N /stop. Insets show the total activity of the major AChE forms, which were deduced from the sedimentation profiles. In the case of high doses of Q N /H C , we calculated the amount of G 1 na by assuming that the proportion of G 2 a to G 1 a remained constant, as in the case of Q N /stop. therefore represents a true secretory process and does not result from cell lysis or other damage. These modifications may include maturation of glycans, proteolysis, or possibly palmitoylation (27). When a flag peptidic epitope was added by mutagenesis at the COOH terminus of the rat AChE T subunit, it did not modify the production, oligomeric assembly, or secretion of the enzyme. The epitope could be detected by a specific monoclonal antibody (M2) in the cellular but not in the secreted molecules, indicating that secretion could be accompanied by proteolytic cleavage. If cleavage does occur within the T peptide, however, only few residues are removed, since antibodies raised against its last 10 amino acids recognize both cellular and secreted molecules. This is consistent with the fact that the secreted monomers and dimers retain their amphiphilic character, which probably depends on the presence of an amphiphilic ␣-helix, constituted by the first 20 amino acids of the T peptide. 2 Note, however, that Liao et al. (17) observed that soluble monomers from bovine brain did not react with their anti-COOH-terminal peptide. We also found that media recovered after several days of culture could contain nonamphiphilic monomers (G 1 na ), probably resulting from proteolysis of the T peptide, as already observed by Velan et al. (28), in the case of HEK 293 cells expressing the human AChE T subunit. Taken together, these observations suggest the existence of multiple cleavage sites. It will be interesting to investigate whether secreted dimers (G 2 a ) conserve the cysteine residue that is located at position Ϫ4 from the COOH terminus and forms an intersubunit disulfide bond in tetramers.
Cells expressing Torpedo AChE T subunits produced the G 2 a form but no G 1 a form and released very little AChE activity. This difference could not be accounted for by the fact that the cultures were maintained at 37°C in the case of rat AChE, but transferred to 27°C to produce Torpedo AChE in an active form. When rat AChE was expressed at 27°C, the total activity was reduced to about one-third of its value at 37°C, but the ratio of secreted to cellular enzyme was approximately the same (not shown). Moreover, the heteromeric G 4 na form of Torpedo AChE, obtained in coexpression with a Q N /stop protein, was secreted readily at 27°C. It seems, therefore, that the Torpedo G 2 a form cannot be transported efficiently to the membrane and released into the medium.
The structure of AChE T is remarkable because these various oligomeric states of the enzyme are not in equilibrium; extracted or secreted molecules may form aggregates under spe-cific conditions but were never observed to assemble into well defined oligomers such as those produced in the cells. The nature of the enzyme has a crucial influence on the proportions of oligomeric and monomeric forms produced in cells; this depends on the alternative COOH-terminal peptides as shown by the difference between rat AChE T and AChE H in rat basoleukemia cells (17), but also on the species. Thus, Torpedo AChE preferentially produces dimers, and rat AChE mostly produces monomers. These proportions also depend on the nature and state of differentiation of the cell; for example, it varies during development of the nervous system (for review, see Ref. 1), and AChE T subunits are expressed in a tissue-specific manner in Xenopus embryos, even under the control of a viral promoter (29,30). In culture, the yield of AChE activity was markedly less in rat basoleukemia cells than in COS cells, in parallel transient transfections (17). We show here that, depending on their state, the COS cells themselves produce variable patterns of molecular forms from AChE T subunits. In the case of the rat enzyme, the cells may contain almost exclusively G 1 a , or equal amounts of G 1 a and G 2 a , together with significant proportions of G 4 a , G 4 na , and 13.7 S oligomers. Thus, the cellular environment somehow controls the formation of oligomers. domain associated with a GPI addition signal (Q N /H C ), produced heteromeric molecules, G 4 na or GPI-G 4 a . In the case of Torpedo AChE, this allowed an active exportation and release of tetramers, whereas dimers were almost entirely retained within the cells. Immunofluorescence of permeabilized cells showed that the enzyme appeared distributed in the reticulum when expressed alone, but concentrated into subcellular bodies, perhaps secretory vesicles, when coexpressed with Q N /H C .

Interaction between the Q N Domain and AChE T in Cotrans
At low doses of DNA encoding the binding proteins, we observed an increase in the production of total AChE (cellular and secreted), in agreement with the incorporation of monomeric G 1 a and dimeric G 2 a forms, which present a rapid metabolic turnover rate (20) into more stable tetramers. However, the total yield of AChE was decreased at high doses of DNA encoding Q N /stop and more markedly Q N /H C , probably because of competition between the production of catalytic and structural subunits.
The heteromeric G 4 na molecules obtained by association of Q N /stop with AChE T tetramers were soluble and secreted, resulting in a high yield of AChE activity in the culture medium. In contrast, the GPI-anchored molecules obtained by expression of Q N /H C alone or in combination with AChE T were mostly attached to the cellular surface, as shown by immunofluorescence of intact cells. Thus, the presence of Q N /H C induced the assembly of GPI-anchored tetramers of AChE T . Coexpression with Q N /H C also produced some lytic G 4 na ; in addition, at high doses of DNA encoding Q N /H C , and in some batches of COS cells, we also observed the production of G 1 na molecules, both in cell extracts and in the medium. Since this G 1 na form was not observed in parallel cotransfections with Q N /stop, the production of these lytic molecules possibly occurs in the metabolic pathway of GPI-anchored proteins, e.g. after reinternalization from the cell surface. Note, however, that cells expressing only AChE T also produced a G 1 na form after several days in culture, as also observed in HEK 293 cells (28), but this may result from another type of cleavage.
We considered the possibility that a Q N domain might induce the formation of AChE T tetramers, without necessarily participating in the final quaternary structure. This hypothesis seemed consistent with the fact that homotetramers are formed at a low level, in the absence of binding proteins, and that the pair of vicinal cysteines of the Q N domain could be mutated without affecting the production of G 4 molecules (31). To detect the presence of Q N in G 4 molecules, we used an antiserum raised against this domain. We found that the sedimentation of released G 4 na was shifted in the presence of this antiserum to the same extent as that of PI-PLC-treated molecules derived from GPI-anchored G 4 a , indicating that Q N was associated with the released G 4 na tetramers. The presence of Q N in the secreted soluble G 4 na form, as well as in the GPI-anchored cellular G 4 a form, was confirmed by inserting a flag epitope in NH 2 -terminal position (N-flagged Q N /stop and Q N /H C ), and in COOHterminal position (Q N /flag). The M1 monoclonal antibody reacted only with a fraction of heteromeric tetramers when a flag epitope was introduced between residues 42 and 43 of the precursor of Q N (both in Q N /stop and Q N /H C ), possibly because it could be cleaved during maturation. Although limited, this reaction indicated that cleavage of the signal peptide occurs at the indicated position, that the mature protein probably starts at Glu-43, and that the flag peptide does not prevent interaction with AChE T . In this study we showed that AChE T subunits, when ex-pressed in COS cells, are able to generate a number of various more or less stable oligomeric associations. Coexpression with binding proteins derived from the NH 2 -terminal domain of the collagenic tail induces the preferential formation of heteromeric molecules in which AChE T tetramers are assembled with a binding domain. This appears to facilitate the transport of AChE T subunits in the secretory pathway and exportation to the cell surface, in the case of GPI-anchored molecules, or their secretion, in the case of soluble molecules. The recruitment of AChE T subunits depends on the dose of binding protein, in a saturable manner. It is therefore possible to analyze these protein-protein interactions quantitatively in the secretory pathway of living cells.