Quaternary associations of acetylcholinesterase. II. The polyproline attachment domain of the collagen tail.

In transfected COS cells, we analyzed the formation of heteromeric associations between rat acetylcholinesterase of type T (AChET) and various constructions derived from the NH2-terminal region of the collagen tail of asymmetric forms, QN. Using a series of deletions and point mutations in QN, we showed that the binding of AChET to QN does not require the cysteines that normally establish intersubunit disulfide bonds with catalytic subunits and that it essentially relies on the presence of stretches of successive prolines, although adjacent residues also contribute to the interaction. We thus defined a roline-ich ttachment omain or PRAD, which recruits AChET subunits to form heteromeric associations. Such molecules, consisting of one PRAD associated with a tetramer of AChET, are exported efficiently by the cells. Using the proportion of AChET subunits engaged in heteromeric tetramers, we ranked the interaction efficiency of various constructions. From these experiments we evaluated the contribution of different elements of the PRAD to the quaternary assembly of AChET subunits in the secretory pathway. The PRAD remained functional when reduced to six residues followed by a string of 10 prolines (Glu-Ser-Thr-Gly3-Pro10). We then showed that synthetic polyproline itself can associate with AChET subunits, producing well defined tetramers, when added to live transfected cells or even to cell extracts. This is the first example of an in vitro assembly of AChE tetramers from monomers and dimers. These results open the way to a chemical-physical exploration of the formation of these quaternary associations, both in the secretory pathway and in vitro.

In transfected COS cells, we analyzed the formation of heteromeric associations between rat acetylcholinesterase of type T (AChE T ) and various constructions derived from the NH 2 -terminal region of the collagen tail of asymmetric forms, Q N . Using a series of deletions and point mutations in Q N , we showed that the binding of AChE T to Q N does not require the cysteines that normally establish intersubunit disulfide bonds with catalytic subunits and that it essentially relies on the presence of stretches of successive prolines, although adjacent residues also contribute to the interaction. We thus defined a proline-rich attachment domain or PRAD, which recruits AChE T subunits to form heteromeric associations. Such molecules, consisting of one PRAD associated with a tetramer of AChE T , are exported efficiently by the cells. Using the proportion of AChE T subunits engaged in heteromeric tetramers, we ranked the interaction efficiency of various constructions. From these experiments we evaluated the contribution of different elements of the PRAD to the quaternary assembly of AChE T subunits in the secretory pathway. The PRAD remained functional when reduced to six residues followed by a string of 10 prolines (Glu-Ser-Thr-Gly 3 -Pro 10 ). We then showed that synthetic polyproline itself can associate with AChE T subunits, producing well defined tetramers, when added to live transfected cells or even to cell extracts. This is the first example of an in vitro assembly of AChE tetramers from monomers and dimers. These results open the way to a chemical-physical exploration of the formation of these quaternary associations, both in the secretory pathway and in vitro.
As shown in the preceding article (1), coexpression in COS cells offers a convenient method to analyze the formation of heteromeric molecules in which tetramers of acetylcholinesterase (AChE, 1  When expressed alone in COS cells, AChE subunits of type T (AChE T ) produce mostly monomers and dimers, with smaller proportions of tetramers and higher oligomers. In the presence of the binding proteins Q N /stop and Q N /H C , these subunits are recruited into tetramers that are associated with the binding domain and are efficiently secreted or attached to the cell surface by a glycolipidic anchor (GPI) (1).
In this study we used a series of deletions and point mutations in the Q N sequence in an attempt to define the attachment domain that allows its interaction with rat AChE T . We show that the cysteine residues are not required for interaction with AChE T , and we narrow down the binding domain essentially to a stretch of successive prolines. Moreover, we show that polyproline itself can combine with AChE T subunits, inducing their polymerization, mainly into tetramers, when added to transfected cells or to a cell extract.

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). Poly-L-proline with a molecular mass of 1,000 -10,000 (mean 8 kDa) and Ͼ30,000 (mean 40 kDa) was purchased from Sigma.
Site-directed Mutagenesis and Transfection in COS Cells-Expression vectors encoding the binding proteins Q N /H C and Q N /stop (Q N / stop551) were described previously (1,3). The structure of the proteins is illustrated in Fig. 1. Site-directed mutagenesis was performed with the single strand method (4). In the case of deletions, we used mutagenic 20-mer oligonucleotides consisting of 10 nucleotides complementary to each side of the deleted fragment. In the case of truncations, TGA stop codons were introduced to terminate the polypeptide chain. Transfection and culture of COS cells were performed as described in the preceding paper (1).
Analysis of AChE Forms-PI-PLC treatments, sedimentation and electrophoretic analyses, were performed as described (1).

Definition of the Proline-rich Attachment Domain (PRAD)
Deletions in the COOH-terminal Part of the Q N Domain-We examined the effect of deletions in the COOH-terminal part of the Q N domain on the binding of AChE T subunits. For this purpose we used two strategies: introduction of stop codons at various positions in the Q sequence, producing Q N /stop proteins of various lengths; or deletions of various extents in the chimeric Q N /H C protein (Fig. 1).
When coexpressed with rat AChE T subunits, Q N molecules truncated at position 87 (Q N /stop87) induced a significant increase in the proportion of cellular G 4 na form and more dramatically, a large increase in the released activity, where G 4 na became predominant over G 2 a and G 1 a (not shown). Thus, Q N / stop87 was able to induce the formation of heteromeric tetramers, like the Q N /stop551 protein, containing the complete Q N domain, which we analyzed in the preceding paper (1). In contrast, introduction of a stop codon at position 76 totally abolished this capacity. In the case of Q N /stop76, the pattern of * This research was supported by grants from the CNRS, the Direction des Recherches et Etudes Techniques, the Association Française contre les Myopathies, and the Human Capital and Mobility program of the European Community. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. AChE forms was identical to that of AChE T subunits alone, as illustrated in nondenaturing electrophoresis (Fig. 2). We performed similar experiments in cotransfections with Torpedo AChE T subunits; the cells released essentially no activity when Torpedo AChE T was expressed alone or with Q N /stop76, but coexpression with Q N /stop87 induced the secretion of tetramers in the culture medium (not shown).
Deletions in the Q N /H C molecule are schematically shown in Fig. 1B. Deletion of COOH-terminal segments of the Q N domain, extending up to positions 100 or 85 (Q N ⌬100 -110/H C , Q N ⌬85-110/H C ) did not abolish the production of a GPI-G 4 a form, as shown by the sedimentation profiles of the cellular enzyme and its sensitivity to PI-PLC (not shown). However, Q N ⌬85-110/H C somewhat weakened this interaction (see below). In contrast, larger deletions extending into the prolinerich sequence that follows the two vicinal cysteines of Q N (Q N ⌬81-110/H C and Q N ⌬76 -110/H C ) abolished the binding of AChE T .
Together these deletion experiments showed that a large part of the COOH-terminal sequence of Q N may be removed without compromising the binding of AChE T subunits from rat or from Torpedo. They defined a boundary of the binding domain, between positions 81 (essentially no binding) and 87 (similar to the wild type). NH 2 -terminal and COOH-terminal Deletions in the Q N Domain-To assess the possible role of the peptidic sequence that precedes the pair of cysteines Cys 70 -Cys 71 in the Q N domain, we removed residues 46 -69 in deletion ⌬1, leaving only the three residues that immediately follow the putative cleavage site of the signal peptide. In deletion ⌬2, we deleted residues 90 -110 in the COOH-terminal part of Q N . We introduced these deletions, separately and in combination, in the Q N /H C and Q N /stop proteins. All of these deleted molecules were able to combine with the rat AChE T subunit. For example, Q N /stop551 proteins carrying one or both deletions induced the secretion of G 4 na rather than G 1 a or G 2 a (not shown). In the case of Q N /H C , the production of GPI-anchored G 4 a and the release of G 4 na were similar for the three deleted constructs and for the complete protein, as shown by nondenaturing electrophoresis (Fig. 3) and by sedimentation analyses (not shown). Surprisingly, the ⌬1 deletion actually increased the interaction with AChE T , at nonsaturating doses of DNA encoding Q N /H C or Q N /stop (see below).
These observations establish that at least part of the sequence located between residues 70 and 86 plays a critical role in the interaction of Q N with the AChE T catalytic subunits. The most prominent feature of this short peptidic sequence is the presence of two stretches of five and three prolines, separated by two residues, Met 80 -Phe 81 .
Are Disulfide Bonds Necessary for Association of Q N with AChE T Subunits?-The Q N domain contains a pair of vicinal cysteines (Cys 70 -Cys 71 ), which form disulfide bonds with cysteines located at position Ϫ4 of the COOH terminus of two AChE T subunits, whereas the same cysteine residues form a disulfide bond between the other two subunits of the catalytic tetramer (11). To examine whether the formation of disulfide bonds between Q N and AChE T subunits was a necessary requirement for the assembly or stability of the hetero-oligomeric structure, we constructed mutants that lacked the vicinal cysteines. The double mutation C70G/C71S was introduced in the entire Q subunit and in the Q N /H C chimeric protein. In both Cellular extracts and culture media of cells expressing rat AChE T together with Q N /stop76 or Q N /stop87 were analyzed by nondenaturing electrophoresis, as indicated. The patterns obtained with Q N /stop76 were identical to those obtained with AChE T alone; Q N /stop87 induced a secretion of G 4 na instead of G 1 a and G 2 a .
cases, the mutated proteins behaved exactly like those containing the pair of cysteines, producing hetero-oligomers with rat AChE T subunits: collagen-tailed forms in the case of Q, GPI-anchored and released tetramers in the case of Q N /H C (not shown). It was possible to obtain a more quantitative comparison, using the proportion of heteromeric AChE to total activity as an index of the efficiency of interaction, if we assume that the different constructs were expressed at the same level. (Note that they all contained the same signal peptide and cleavage site.) We thus show a ladder of the different Q N /stop constructions, according to the proportion of G 4 na in the culture medium (Fig. 4). This shows that Q N /stop85 was slightly but significantly less efficient than Q N /stop87, which was itself less efficient than Q N /stop551. The Phe 85 and Phe 86 residues therefore contribute to the binding of AChE T subunits. The effect of the ⌬2(90 -533) deletion also suggests that more distal residues participate in this interaction. In the case of the Q N /H C constructs, it is possible that the interaction may depend on a sufficient distance between the binding domain and the GPI addition signal.

Mutations in the Proline-rich Region; Comparisons of the Efficiency of Interaction with AChE T ; Construction of a Mini
To assess the possible importance of Met 80 and Phe 81 , located between the two proline stretches, in the interaction with AChE T subunits, we mutated these residues into prolines. When coexpressed with rat AChE T subunits, the Q N [Pro 80 -Pro 81 ]/H C mutant was able to produce GPI-anchored G 4 a , as well as released G 4 na , with a slightly reduced efficiency compared to the wild type Q N /H C protein (not shown). Similarly, the Q N [Pro 80 -Pro 81 ]/stop85, Q N [Pro 80 -Pro 81 ]/stop87, and Q N [Gly 72 -Gly 73 -Gly 74 -Pro 80 -Pro 81 ]/stop85 mutants produced secreted G 4 na less efficiently than the corresponding constructions containing Met 80 -Phe 81 (Fig. 4).
We also examined the importance of the proline stretches by replacing the middle residue of each group by a glycine. We found that whereas the P83G mutation had little effect on the yield of GPI-anchored G 4 a and released G 4 na , the P77G mutation reduced it markedly, and the double mutation P77G/P83G further weakened the binding. Introducing prolines at positions 80 and 81 partially compensated the effect of P77G but weakened the interaction in the case of P83G or of the double mutation P77G/P83G.
The two cysteines of the attachment domain are separated from the proline stretches by three residues, Leu 72 -Leu 73 -Thr 74 . Mutation of these residues to glycines in complete or partially deleted Q N /H C was found to weaken significantly the interaction with AChE T . In agreement with the effects of mutations of the Cys 70 -Cys 71 and Leu 72 -Leu 73 -Thr 74 residues, a deletion including the two cysteines (⌬46 -71) was as effective as deletion ⌬1(46 -69) in which they were maintained, whereas removal of the following Leu 72 -Leu 73 residues significantly reduced the efficiency of interaction (Fig. 4).
These observations indicate that a sufficient stretch of proline residues is essential for the association with AChE T but that hydrophobic residues located before, between, and after the two proline stretches (Leu 72 -Leu 73 -Thr 74 , Met 80 -Phe 81 , Phe 85 -Phe 86 ) also participate in this interaction. We constructed a mutant, Q N /⌬46 -74Pro 80 -Pro 81 /stop85, in which the mature protein only consisted of three residues, EST, followed by a stretch of 10 prolines, Pro 10 , as in the Pro 80 -Pro 81 -mutant. We found that this Glu-Ser-Thr-Pro 10 peptide was able to combine with AChE T , decreasing the proportion of G 2 a and G 1 a and increasing markedly the level of secreted G 4 na , although with less efficiency than other Q N /stop constructs (Fig. 4). This suggested that the polyproline sequence may be sufficient to organize AChE T tetramers in the absence of other residues.
Conversely, Q N /H C and Q N /stop constructs in which residues 70 -86 were deleted had no effect on the molecular forms of AChE, as shown by sedimentation analyses or by immunofluorescence (not shown). They did not induce the production of GPI-anchored G 4 a at the surface of the cells or the secretion of G 4 na in the medium.

Interaction of AChE T Subunits with Synthetic Polyproline
Exogenous, Synthetic Polyproline Can Combine with AChE T in Transfected Cells-We added various concentrations of synthetic polyproline of about 8 kDa to the culture medium after transfecting cells with AChE T . We observed a dose-dependent decrease of G 2 a and G 1 a , both in the cell extract (Fig. 5A) and in the culture medium (Fig. 5B), and a concomitant increase of G 4 na . These effects could be detected at 5 ϫ 10 Ϫ8 M polyproline; at higher concentrations G 4 na became the major secreted form of AChE, as observed with Q N /stop constructs. In addition, a minor 16 S component appeared in the cell extract and in the medium at higher concentrations of polyproline. 2 The sedimentation coefficient of this form suggests that it may result from the binding of two tetramers to a sufficiently long polyproline chain.
Polyproline of higher average molecular mass (40 kDa) also interacted with AChE but induced the formation of ill defined components, sedimenting mainly between 6 and 11 S.
We wondered whether AChE T and polyproline could combine spontaneously or whether the cellular biosynthetic machinery was required for this interaction. We found that ␣␣Ј-dipyridyl, which inhibits hydroxylation of prolines, did not inhibit the production of G 4 na by polyproline, even when included at 10 Ϫ4 M in the culture medium for 3 days after transfection. This concentration reduced the production of total AChE activity to less than 20% of the control but did not alter the proportions of molecular forms (not shown).
Interaction of Polyproline with AChE T in Cell Extracts-We also examined whether the interaction could occur in an acellular system. When a cell extract, obtained in the absence of detergent (high speed supernatant of a low salt-soluble extract), was incubated with polyproline at 37°C for 4 h, we observed an important increase of G 4 na at the expense of G 1 a and G 2 a (Fig. 6A). The effect of polyproline on AChE monomers and dimers in a detergent-free cell extract resembled the effect obtained with living cells, except for the absence of the 16 S component. The 13.7 S form was not observed under these conditions, probably because it disappeared during incubation at 37°C. The efficiency of interaction was reduced markedly in the presence of Triton X-100 (Fig. 6B). In addition, the production of G 4 na , in the presence of polyproline, occurred less efficiently or more slowly at lower temperatures, 20 or 4°C (not shown). The interaction between AChE and polyproline was therefore dose-dependent, temperature-dependent, and sensitive to the presence of detergent.
In contrast with cell extracts, we observed only a minimal effect after incubation of the secreted AChE forms with polyproline (less than 5% of G 1 a and G 2 a was converted into G 4 na ), and this could in fact reflect the contribution of enzyme forms released by cell lysis. Experiments in which cellular extracts and media from transfected and control COS cells were combined, so that the composition of the mixtures was identical except for the cellular or secreted origin of the AChE molecules, showed that the difference in their capacity to interact with polyproline is an intrinsic property of these monomers and dimers (not shown). DISCUSSION In these experiments, we used a series of deletions and mutations in the NH 2 -terminal domain of the collagen Q sub- unit to identify the residues that are involved in the attachment of a tetramer of AChE T subunits. These modifications were introduced in the Q N /H C and Q N /stop proteins and in some cases also in the complete collagenic subunit, Q. Assembly of AChE T subunits with Q is less convenient because it produces several collagen-tailed molecules (A 12 , A 8 , A 4 ) which aggregate in low salt and cannot be analyzed in nondenaturing electrophoresis. The results were however entirely consistent with those obtained for Q N /H C and Q N /stop proteins and have not been illustrated here.
Cysteines Are Not Required for Interaction of Q N with AChE T -By comparing the proportions of heteromeric tetramers we were able to rank the efficiency of interaction of different deleted and mutated constructions and thus evaluate the contribution of various elements of the Q N domain. Quite unexpectedly, deletion of the peptide preceding the cysteines (⌬1) actually increased the interaction significantly. We also showed that the vicinal cysteines of Q N , Cys 70 and Cys 71 , are not required, although disulfide bonds probably stabilize the quaternary association with AChE T . The fact that Cys 70 and Cys 71 are dispensable is consistent with our previous observation that intersubunit disulfide bonds may be reduced without disrupting the structure of collagen-tailed AChE forms (5). We also showed that sequences following a polyproline stretch, i.e. beyond position Phe 86 , could be removed without compromising the binding capacity.
PRAD, a Short Conserved Peptidic Domain; Evaluation of Interaction Efficiency in the Secretory Pathway-These experiments showed that the binding domain could be reduced to a short region of 17 residues, starting with cysteines Cys 70 -Cys 71 , which we propose to name the polyproline attachment domain or PRAD. This region is remarkably well conserved from Torpedo to higher vertebrates; there are only two replacements in the rat Q subunit 3 and none in the chicken. 4 The presence of PRAD is sufficient for interaction with AChE T subunits, and conversely its deletion abolishes this interaction completely.
The PRAD contains two conserved stretches of five and three consecutive prolines, separated by two residues, Met 80 -Phe 81 , preceded and followed by conserved residues, Leu 72 -Leu 73 -Thr 74 and Phe 85 -Phe 86 . By point mutations and deletions we showed that these hydrophobic residues, as well as more distal residues, participate in the interaction with AChE T subunits but are not absolutely required.
By analyzing the ratio R of secreted heterotetramers, G 4 na , to the total activity in the culture medium for nonsaturating doses of Q N /stop constructions, as described in the previous paper (1), we obtained an index of the efficiency of interaction, assuming that the binding proteins were expressed at the same level (Fig. 4). This index provides an evaluation of the influence of elements of the binding domain; for example, the M80P/F81P double mutation reduces the value of R to a similar degree in different contexts (Q N /stop87, Q N /stop85, or Q N Gly 72 -Gly 73 -Gly 74 /stop85). It seems therefore possible to quantify the contributions of specific residues to quaternary assembly in the secretory pathway.
Interaction of AChE with Synthetic Polyproline-The critical feature of the binding domain PRAD appeared to be the presence of a sufficiently long series of prolines, since a simple Glu-Ser-Thr-Pro 10 peptide remained functional. We showed that, in effect, AChE T subunits were able to combine with synthetic polyproline. When added to cultures of COS cells expressing AChE T , synthetic polyproline of about 8 kDa was able to mimic the effect of the Q N domain, even at a concentra-tion as low as 10 Ϫ9 M, inducing the formation and the release of tetramers G 4 na at the expense of monomers and dimers. It also produced a heavier complex of 16 S, in lesser proportion. Polyproline was probably taken up by endosomes so that a fraction was able to reach the Golgi apparatus and the endoplasmic reticulum and thus to interact with AChE during its biosynthesis and maturation. We showed that ␣␣Ј-dipyridyl, which blocks proline hydroxylation, had no effect on this interaction either in cells or in vitro, even at concentrations that inhibited AChE biosynthesis by more than 70%.
In fact, the biosynthetic machinery of the cell was not absolutely necessary since this interaction could be reproduced in a soluble extract of transfected COS cells expressing rat AChE T . Polyproline induced the formation of tetramers G 4 na from monomers and dimers of rat AChE T and possibly also of dimers from monomers. The effect was dose-dependent and temperature-dependent and was reduced considerably in the presence of Triton X-100. It is likely that interaction of the detergent with the COOH-terminal peptide of AChE T , organized as an amphiphilic ␣-helix 5 prevents its interaction with polyproline. This is the first time that oligomerization of AChE has been obtained in vitro. In previous studies, molecular forms of AChE were dissociated, e.g. by limited proteolysis, or reversibly aggregated, as in the case of collagen-tailed molecules at low ionic strength in the presence of polyanions (6); but monomers or dimers were never seen to form stable tetramers. There was, however, a clear difference between the oligomerization of AChE T subunits obtained in living cells and in acellular extracts, since heavy polymers, sedimenting around 16 S, were obtained only in the cells.
It is worth noting that molecules solubilized from cells were able to interact with polyproline, whereas those secreted in the culture medium were not. This is consistent with the fact that post-translational modifications occur during transport and secretion (1). Cleavage of the COOH-terminal T peptide may affect the ability of monomers to form oligomers. For example, soluble monomers that had been purified from bovine brain were found to lack the COOH-terminal part of the T peptide (7). It should be noted, however, that the secreted G 1 a and G 2 a forms obtained in our culture media retained their amphiphilic character, which reveals the presence of the amphiphilic ␣-helix of the T peptide.
Possible Interaction of PRAD with Other Proteins-The proline-rich structure of PRAD is reminiscent of the peptide sequences that bind to SH 3 domains (8,9). By analogy, it is therefore tempting to suggest that PRAD is organized as a polyproline II helix. There are marked differences, however, between these proline-rich domains. First, the PRAD exists in the secretory pathway, establishing interactions between externalized proteins, whereas SH 3 binding domains are located in the cytosol; such polyproline sequences have not been found previously in extracellular proteins. Second, polyproline in PRAD does not simply provide a scaffold but may actually mimic the interaction. Third, there is no similarity between SH 3 domains and the COOH-terminal T peptides of AChE T , which are necessary for attachment to the collagen tail. Finally, the stoichiometry of the binding is different, since PRAD simultaneously binds to four AChE T subunits. This in fact raises an interesting question about the symmetry of the heteromeric assembly, since polyproline II has an helical repetition of three residues per turn. In any case, it will be interesting to investigate the possible interaction of PRAD with other proteins, such as the 100-kDa protein that is associated with Torpedo asymmetric forms (10,11). It will also be interesting to examine the possible existence of PRAD in other modular ex-tracellular molecules apart from the Q collagen.
In conclusion, the present study reveals a new type of interaction that occurs intracellularly and withstands extracellular conditions. Such quaternary interactions are important for anchoring AChE and possibly other enzymes and proteins in extracellular matrices. We show that such interactions can induce molecular assembly both in the secretory pathway of living cells and in vitro.