Two Different Heparin-binding Domains in the Triple-helical Domain of ColQ, the Collagen Tail Subunit of Synaptic Acetylcholinesterase*

ColQ, the collagen tail subunit of asymmetric acetylcholinesterase, is responsible for anchoring the enzyme at the vertebrate synaptic basal lamina by interacting with heparan sulfate proteoglycans. To get insights about this function, the interaction of ColQ with heparin was analyzed. For this, heparin affinity chromatography of the complete oligomeric enzyme carrying different mutations in ColQ was performed. Results demonstrate that only the two predicted heparin-binding domains present in the collagen domain of ColQ are responsible for heparin interaction. Despite their similarity in basic charge distribution, each heparin-binding domain had different affinity for heparin. This difference is not solely determined by the number or nature of the basic residues conforming each site, but rather depends critically on local structural features of the triple helix, which can be influenced even by distant regions within ColQ. Thus, ColQ possesses two heparin-binding domains with different properties that may have non-redundant functions. We hypothesize that these binding sites coordinate acetylcholinesterase positioning within the organized architecture of the neuromuscular junction basal lamina.

At vertebrate cholinergic synapses, acetylcholinesterase (AChE) 1 rapidly hydrolyzes the neurotransmitter acetylcholine, thereby terminating synaptic transmission. This key function does not only require a high catalytic turnover number but also a strategic positioning of the enzyme. This is achieved by the association of AChE catalytic subunits with structural subunits that bring them to the site of action. In the case of asymmetric AChE, the collagen ColQ is responsible for the localization of the enzyme at the vertebrate neuromuscular junction. Inactivation of the ColQ gene in mice or mutations in the human ColQ gene result in the absence of enzyme accumu-lation at the neuromuscular junction and are the cause of a congenital myasthenic syndrome (type 1c) (1,2).
As found in other collagens, ColQ contains a central triplehelical domain surrounded by non-collagenous N-and C-terminal domains (Fig. 1A). Each N-terminal domain organizes an AChE tetramer, so the triple-helical structure generates an A 12 or asymmetric AChE form with 12 catalytic subunits (3,4). The collagen domain is characterized by Gly-Xaa-Yaa repeats and a high proportion of the stabilizing proline and hydroxyproline residues. The C-terminal domain is divided into a proline-rich region, important for triple-helix formation (5), and a cysteinerich region probably involved in the anchorage of AChE at the neuromuscular junction, since mutations in this region prevent the accumulation of AChE in congenital myasthenic syndrome type 1c patients (2).
Heparan sulfate proteoglycans (HSPGs) have been implicated in the anchorage of asymmetric AChE to the synaptic basal lamina by interacting with ColQ through their heparan sulfate (HS) chains. This was proposed after showing that AChE could be specifically solubilized from the tissue with heparinase as well as with HS and heparin (6). Consistently, binding of the enzyme to the cell surface is inhibited after treatment of the myotubes with heparitinase and in cells deficient in HS synthesis (7). Later studies showed that exogenously added asymmetric AChE associated specifically with nerve-muscle contact sites in a heparin-sensitive manner (8), suggesting that HSPGs would define its localization at the neuromuscular junction. Perlecan, a basal lamina HSPG, has been proposed as the receptor of collagen-tailed AChE (9), consistent with the recent finding that AChE is not accumulated at the neuromuscular junction in perlecan-null mice (10).
In this context, we have found that the collagen domain of ColQ contains two heparin-binding consensus sequences of the BBXB form (where B represents a basic residue) surrounded by other basic residues and proposed they constitute the sites for HSPG recognition (11). We have characterized the structural properties of those putative heparin-binding domains (HBDs) by using molecular modeling and synthetic peptides (12)(13)(14). Here, we analyzed ColQ-heparin interactions by heparin affinity chromatography of the entire A 12 AChE, carrying different mutations in ColQ. We first established that recombinant and purified A 12 enzymes share the same properties. Then we dissected the participation of the different domains of ColQ as well as of the individual basic residues composing both HBDs to define the determinants of heparin binding. The results obtained provide evidence that expands the role of ColQ-HSPGs interactions to the fine ultrastructural organization of AChE at the synaptic basal lamina. Moreover, they contribute to our understanding of the nature of a HBD in the context of the canonical structure of collagen.

EXPERIMENTAL PROCEDURES
Materials-All restriction endonucleases were purchased from New England Biolabs (Ozyme, France). Other molecular biology reagents were from Promega Corp. (Madison, WI). Oligonucleotides were synthesized either by Oligo Express or Eurobio (Paris, France). Heparinagarose was purchased from Pierce. Heparin-albumin was obtained from Sigma. Acridine-agarose was kindly provided by Dr. Terry Rosenberry (Mayo Clinic, Jacksonville, FL). Unless otherwise specified, other reagents were obtained from commercial sources.
Site-directed Mutagenesis-Two different plasmids were constructed using the coding sequences for rat AChE and ColQ1a (15,16). Both coding sequences were inserted between 5Ј-and 3Ј-untranslated sequences of Xenopus globin mRNA and then subcloned into the pcDNA3 vector (Invitrogen). The recombinant vector pcDNA3-ColQ1a served as a template for site-directed mutagenesis, performed as described by Kunkel (17). To confirm the presence of desired mutations, all mutant plasmids were sequenced by Génome Express (Montreuil sous Bois, France).
Expression in Xenopus Oocytes-Capped synthetic transcripts were prepared using the Ambion mMESSAGE mMACHINE TM in vitro transcription kit (Austin, TX). Samples of ϳ50 nl containing 2.5 ng of AChE mRNA and different amounts of ColQ mRNA were injected with a Drummond Nanojet injection system (Broomall, PA) into the animal poles of Xenopus oocytes (18). Injected oocytes were incubated at 18°C for 24 -48 h in Barth medium (5 mM HEPES, pH 7.6, 98 mM NaCl, 1.8 mM CaCl 2 , 1 mM MgCl 2 , and 50 g/ml gentamycin) supplemented with 50 g/ml ascorbic acid to promote triple-helix formation.
Extraction and Isolation of Recombinant AChE-Injected Xenopus oocytes were homogenized by repeated pipetting in 10 l/oocyte of ice-cold extraction buffer containing 50 mM Tris-HCl, pH 7.0, 10 mM MgCl 2 , 0.8 M NaCl, 1% Brij-96. Different AChE molecular forms were isolated by velocity sedimentation in 5-20% linear sucrose gradients containing extraction buffer. Sucrose gradients were centrifuged at 40,000 rpm for 15 h at 7°C in a SW41 Beckman rotor (Fullerton, CA) and were collected in ϳ48 fractions of ϳ320 l each from which 10 l/fraction were used to measure AChE activity (19). The sedimentation coefficients for AChE forms were determined from those of ␤-galactosidase (16 S) and alkaline phosphatase (6.1 S) from Escherichia coli, run in the gradients as internal markers. Fractions containing the different AChE forms were pooled and stored at Ϫ80°C for future analysis. For comparison, the different forms of AChE isolated from the sucrose gradient were further purified by affinity chromatography using an acridine-agarose column as for enzyme extracted from Torpedo electric organ, but since they showed exactly the same behavior as the enzyme isolated directly from the sucrose gradients, this enzyme was more often used for the experiments.
Purification of Torpedo Asymmetric AChE-Asymmetric AChE was purified from Torpedo californica electric organ (Pacific Bio-Marine Laboratory, Venice, CA) using sequential extraction and affinity chromatography, as described previously (13).
Heparin-Agarose Affinity Columns-Relative affinity for heparin was analyzed for different ColQ mutants using heparin-agarose affinity columns, where the bound enzyme was eluted with a linear NaCl gradient. After packing 500 l of heparin-agarose in a column without flow, the affinity resin was equilibrated with 50 mM Tris-HCl, pH 7.4, and 100 mM NaCl, and 20 -60 milliunits of asymmetric AChE in equilibration buffer were loaded. The column was washed with 10 ml of equilibration buffer, and the enzyme was finally eluted using a 20-ml linear NaCl gradient of either 0.1-0.8 M NaCl or 0.2-0.9 M NaCl in 50 mM Tris-HCl, pH 7.4. Loading, washing, and elution steps were performed at a constant flow rate of 9.75 ml/h, maintained by a Gilson Miniplus 2 peristaltic pump. Approximately 77 fractions of ϳ260 l each were collected. For all of these, 100-l aliquots were used to determine AChE activity, whereas aliquots of 100 l diluted with 1 ml of distilled water were used to measure conductivity using an Amber Science Inc. 605 conductivity meter (Eugene, OR). Conductivity values were converted to NaCl concentrations using a 5-point standard curve with NaCl solutions made up with elution buffer. NaCl concentrations increased by 0.0105 Ϯ 0.0006 M/fraction (n ϭ 97).
Equilibrium Binding Assays-Binding assays were performed in MaxiSorp 96-well plates (Nunc, Roskilde, Denmark) coated with heparin coupled to albumin as described previously (13). After coating and blocking steps, the plates were incubated with 0.25 milliunits of recombinant A 12 AChE (non-saturating amount of enzyme) in 50 l of 50 mM Tris-HCl, pH 7.4, and 150 mM NaCl for 18 h at 4°C with constant agitation. The supernatant was recovered, the plates were washed twice with 100 l of incubation buffer for 5 min, and AChE activity was measured in the supernatant and washes (both representing the unbound enzyme) as well directly in the plates (bound enzyme).
Data Treatment-To analyze the peak composition of elution profiles of AChE from heparin-agarose columns, the noise present in raw data was filtered using the Savitzky-Golay algorithm, the base line was detected and subtracted using a non-parametric algorithm, and finally, the peaks were detected and fitted to Gaussian peaks using the method of the second derivative (r 2 Ͼ 0.999) using the program PeakFit TM 4.06 from SPSS Inc. (Chicago, IL).
To facilitate the analysis of elution profiles of AChE from heparinagarose and the comparison between different asymmetric AChE mutants, raw data from AChE elution profiles were plotted for AChE activity versus NaCl concentration, fitted to a non-parametric base line that was then subtracted, and converted to cumulative areas using the program PeakFit TM 4.06 from SPSS Inc. (Chicago, IL). Using the Origin 6.0 program from Microcal Software Inc. (Northampton, MA), cumulative area values were normalized to percentages and the data plotted as cumulative areas (%) versus logarithm of NaCl concentration (M) were fitted to a dose-response (or four parameter) sigmoidal equation with variable slope, where the parameters A 1 , A 2 , log x 0 , and p correspond to the bottom asymptote, the top asymptote, the center, and the variable Hill slope, respectively. Finally, starting from the center parameter value (i.e. log[NaCl]), the NaCl concentration for 50% elution of the bound enzyme was obtained for each curve. This value was used in this study as a measure of the relative affinity of a given A 12 AChE mutant for heparin.
In the different experiments, statistical analyses were performed using the unpaired Student's t test.

Production of Recombinant Rat A 12 AChE and Its Affinity for
Heparin-Xenopus oocytes constitute a very efficient system to produce predominantly A 12 AChE forms by injecting defined amounts of both ColQ and AChE mRNAs (16), as shown by the sedimentation profile in a sucrose gradient of oocyte extracts ( Fig. 2A).
To evaluate the heparin-binding behavior of recombinant rat asymmetric AChE, its A 12 form produced in Xenopus oocytes was isolated from a sucrose gradient, bound to a heparinagarose column, and eluted from the resin using a linear NaCl gradient. This strategy allows determination of the values of relative affinity for heparin, where a protein with higher affinity is eluted at a higher NaCl concentration (20). A 12 AChE purified from Torpedo electric organ was also analyzed as reference.
As shown in Fig. 2, B and C, both Torpedo purified and rat recombinant A 12 AChE present the same elution profile, composed of five peaks eluting at the same NaCl concentrations and presenting similar relative abundance. In both cases the majority of the enzyme was eluted in the central peak at a concentration of 0.6 M NaCl. The same elution profile was obtained under different experimental conditions, varying the NaCl concentration increment per fraction, the elution volume, the elution flow rate, and the volume of resin (data not shown). Each peak differed significantly from its neighboring peaks both at the level of NaCl concentration of elution and relative abundance (n ϭ 12, p Ͻ 0.0003). This multiplicity might reflect an intrinsic heterogeneity of enzyme species. To examine this possibility, AChE eluted in the first peak was re-loaded onto a second heparin-agarose column, generating the original five peaks in the elution profile (data not shown). This excludes the possibility that the enzyme is composed of chemically different species. Most likely, the observed behavior arises from the multivalent nature of the HBDs in ColQ (12) and the heterogeneity of heparin molecules.
To facilitate the analysis of the elution profiles of AChE from heparin-agarose and the comparison between different A 12 AChE mutants, the elution profiles were expressed as cumulative areas versus log[NaCl]. The NaCl concentrations at which 50% of the bound enzyme was eluted were derived from these curves (see "Experimental Procedures"), and relate directly to their affinity for heparin. When expressed in this form, both Torpedo and recombinant rat enzymes presented the same elution curve, with 50% of the enzyme eluting at 0.6 M NaCl (Fig. 2D, Table I). These results show that the recombinant rat A 12 AChE behaves exactly as the purified A 12 enzyme from Torpedo electric organ, so that Xenopus oocytes constitute a good expression system for A 12 AChE.
The Heparin-binding Activity Is Localized in the Collagen Domain of ColQ-By analyzing the ColQ sequence, it is possible to identify two heparin-binding consensus sequences of the BBXB type in the collagen domain of ColQ. Other clusters of basic residues also exist within the N-and C-terminal noncollagenous domains of ColQ, and it has been suggested that the C-terminal domain could participate in binding to HSPGs

FIG. 2. Recombinant rat A 12 AChE interacts with heparin as the purified Torpedo enzyme.
A, sedimentation profile of AChE molecular forms produced in Xenopus oocytes. The different peaks from left to right correspond to the asymmetric forms A 12 , A 8 , and A 4 , respectively, and to the globular dimeric (G 2 ) and monomeric (G 1 ) forms. The fractions corresponding to the A 12 peak (16 S) shaded in gray were pooled and used for heparin-binding assays. A schematic representation of the A 12 form is shown. B and C, elution profiles of Torpedo-purified A 12 AChE and recombinant rat A 12 AChE from a heparin-agarose column using a linear NaCl gradient. AChE activity was measured in each fraction (symbols and solid line) as the NaCl concentration (dotted line). Five peaks were detected and fitted by using the method of the second derivative (dashed lines). D, the elution profiles of the Torpedo purified (white circles, from B) and rat recombinant (black circles, from C) enzymes were transformed to cumulative areas, plotted against log [NaCl], and fitted to a four parameter sigmoid (solid lines), allowing determination of the NaCl concentration at which 50% of the heparin-bound enzyme was eluted. This value was equal for both the Torpedo purified and rat recombinant enzymes (p Ͼ 0.7, n ϭ 3). The nomenclature of the mutants was designed to show the state of the N-terminal and the C-terminal HBDs, separated by a slash. When the site is intact it is labeled with a plus sign (e.g. N ϩ ) and when it is inactivated it is labeled with a minus sign (e.g. N Ϫ ). Single and double substitutions are specified. For instance, in N Ϫ /K229A the N-terminal HBD is inactivated, and in the C-terminal HBD lysine 229 was replaced by an alanine. For more complex mutations a conceptual notation is used. For example, in C3 N/C Ϫ the C-terminal HBD is inactivated, and in the N-terminal HBD the basic residues from the N-terminal BBXB motif were replaced by the basic residues from the C-terminal motif, as if the C-terminal site was "moved" to the N-terminal position.
Basic residues are shown in bold, and mutated residues are highlighted in black. The concentration of NaCl required for elution of 50% of the enzyme bound to heparin-agarose columns was obtained from the fitting equation (see "Experimental Procedures") and is expressed in M Ϯ S.D., n varied from 2 to 4; the precise values are given in the figure legends.
(2). To define the contribution of different domains to heparin binding, AChE oligomers containing ColQ with different deletions were expressed and analyzed.
In ⌬Ct, the last 88 residues of ColQ were deleted, corresponding to the cysteine-rich region of the C-terminal domain (Fig.  1B). The remainder of the C-terminal domain was not removed because it is required for triple-helix folding. ⌬Col contains a deletion of 171 amino acids in the central region of the collagen domain of ColQ that eliminates the two HBDs, but 7 prolinerich Gly-Xaa-Yaa triplets were left to ensure the formation of a stable triple helix. Finally, Q N contained only the N-terminal domain of ColQ, allowing the organization of AChE tetramers to occur independently of the collagen domain. Fig. 3A shows the sedimentation profiles of the different ColQ deletion mutants co-expressed with AChE in Xenopus oocytes. As expected, ⌬Ct and ⌬Col were able to produce asymmetric forms of AChE, whereas Q N only produced a tetramer of catalytic subunits (G 4 form). AChE expressed with ⌬Ct presented a sedimentation profile similar to wild type (WT), whereas asymmetric forms produced with ⌬Col presented higher sedimentation coefficients, consistent with a decreased asymmetry of the molecules due to a shortened collagen tail.
The A 12 ⌬Col and ⌬Ct forms as well as the Q N G 4 form were isolated from the sucrose gradients and run on heparin-agarose columns. As shown in Fig. 3B, Q N and ⌬Col completely lost their ability to bind heparin, whereas ⌬Ct presented an elution profile similar to WT (Fig. 3C). The NaCl concentrations at which 50% of the total bound enzyme eluted was 0.61 M for the WT enzyme and 0.63 M for ⌬Ct, not differing significantly (see Table I for mutant sequences, nomenclature, and elution values and Fig. 9 for a summary of relative heparin affinity of all the mutants). Together, these results clearly show that the heparin-binding activity of asymmetric AChE resides exclusively in the collagen domain of ColQ.
The Basic Residues of Both HBDs in the Collagen Domain of ColQ Interact with Heparin-To confirm that the heparin-binding consensus sequences present in the collagen domain of ColQ are responsible for interactions with heparin, double sitedirected mutagenesis was carried out. The first two basic residues of both BBXB consensus sequences were replaced by aspartic acid and proline (N Ϫ /C Ϫ , Table I). Aspartic acid was chosen to obtain a dramatic effect because the opposite charge was likely to cause heparin repulsion. Proline was used as it lacks a reactive side chain and tends to play a structural rather than a functional role (21). Only 55% of N Ϫ /C Ϫ AChE was retained in heparin-agarose, with the bound enzyme eluting at 0.17 M NaCl (Fig. 4, A and B). This suggests that the BBXB sequences in ColQ may be essential for heparin interactions under physiological conditions. Indeed, an equilibrium binding assay performed at 0.15 M NaCl using heparin-coated plates showed that only 16% of the N Ϫ /C Ϫ enzyme was able to bind heparin (Fig. 4A). On the other hand, the fact that N Ϫ /C Ϫ still presented some heparin-binding capacity suggested that the basic residues surrounding the BBXB motif may also participate in this interaction. To assess this possibility, almost all basic residues constituting both putative HBDs were substituted by alanines (N ϭ /C ϭ , Table I). In this case, only 14% of the N ϭ /C ϭ mutant enzyme was retained by heparin-agarose, with only 5% able to bind heparin under equilibrium conditions (Fig.  4A). These results demonstrate that the N-and C-terminal HBDs, comprising BBXB motifs surrounded by other basic residues and localized in the collagen domain of ColQ, are responsible for the interaction of A 12 AChE with heparin.
The Two HBDs in the Collagen Domain of ColQ Have Different Affinities for Heparin-To analyze the behavior of each HBD separately, two ColQ mutants were generated by interrupting a single BBXB motif and leaving the other intact. Thus, N Ϫ /C ϩ allowed the analysis of the C-terminal HBD and N ϩ /C Ϫ of the N-terminal HBD (Fig. 1B). In both cases, the modified BBXB sequence was substituted by DPXB (Table I). Both mutants were fully retained in heparin-agarose as the WT enzyme but eluted from the column at different NaCl concentrations (Fig. 5A). Whereas the N Ϫ /C ϩ enzyme eluted at 0.54 M NaCl, the value for N ϩ /C Ϫ was 0.41 M, implying that the C-terminal HBD has a significantly higher affinity for heparin than the N-terminal HBD. Interestingly, both N Ϫ /C ϩ and N ϩ /C Ϫ mutants presented lower affinities for heparin than the WT enzyme (0.61 M), suggesting that WT affinity results from the simultaneous or cooperative participation of both HBDs.
The Difference between the Two HBDs Does Not Simply Result from the Different Number or Nature of Basic Residues-Although ColQ HBDs share a minimum GRPBBXB sequence, the two sites differ both in the number and nature of basic residues. The C-terminal HBD presents two additional basic residues that are not found in the N-terminal HBD. To evaluate whether these extra basic residues could explain the higher heparin affinity of the C-terminal HBD, these residues were replaced by alanines (N Ϫ /R218A-K233A, Table I). Fig. 5B shows that the resultant mutant had a slightly lower affinity for heparin than N Ϫ /C ϩ , eluting at a concentration of 0.51 M NaCl. This suggests that the extra basic residues at the Cterminal HBD participate in heparin interactions even if their contribution is small. However, N Ϫ /R218A-K233A presented a heparin affinity significantly higher than that of the N-terminal HBD (N ϩ /C Ϫ ), showing that the basic charge difference FIG. 4. Only the two HBDs present in ColQ are responsible for the interaction of AChE with heparin. A, binding of different mutants of A 12 AChE to heparin-agarose columns in 0.1 M NaCl (dark gray bars) or to heparin-coated plates in equilibrium conditions in 0.15 M NaCl (light gray bars). In N Ϫ /C Ϫ , only the two central basic residues of both HBDs were mutated, whereas in N ϭ /C ϭ almost all basic residues were substituted with alanines. Values are the mean Ϯ S.D., n ϭ 3. B, the A 12 AChE mutants retained in the heparin-agarose column were eluted using a linear gradient of NaCl, and the elution profiles were expressed as cumulative areas. The elution values determined from the curves differed significantly between WT and N Ϫ /C Ϫ (p Ͻ 0.0000005, n ϭ 3).

FIG. 5. The two HBDs have different affinity for heparin.
A, cumulative areas of the elution profiles from heparin-agarose columns of the WT A 12 AChE, of N Ϫ /C ϩ , which only presents a C-terminal intact HBD, and of N ϩ /C Ϫ , which contains only the N-terminal HBD. The elution values of both mutants were significantly lower than that of WT (p Ͻ 0.0005 for N Ϫ /C ϩ and p Ͻ 0.00001 for N ϩ /C Ϫ ), and different between them (p Ͻ 0.0001, n ϭ 3). B, elution profiles as cumulative areas of A 12 AChE having only one of the two HBDs, either intact or altered. In N Ϫ /C ϩ and N ϩ /C Ϫ , one of the HBDs has been inactivated, whereas the other remains intact. N Ϫ /N3 C presents only the C-terminal domain intact, in which the BBXB motif has been replaced by the N-terminal sequence. Inversely, C3 N/C Ϫ presents only the N-terminal HBD with the BBXB sequence from the C-terminal HBD. Finally, in N Ϫ /R218A-K233A, the N-terminal HBD has been inactivated, and in the C-terminal HBD the indicated basic residues were replaced by alanines, leaving the central GRPGBBGB sequence intact. See Table I as a guide. N Ϫ /R218A-K233A elution value was significantly lower than that of N Ϫ /C ϩ (p Ͻ 0.0005, n ϭ 3), whereas N Ϫ /C ϩ (n ϭ 3) and N Ϫ /N 3 C (n ϭ 4) had equal elution values (p Ͼ 0.9) as did N ϩ /C Ϫ and C 3 N/C Ϫ (p Ͼ 0.05, n ϭ 3). between the two HBDs does not underlie their different affinities for heparin.
The second difference between the two HBDs lies in the identity of basic residues that constitute each BBXB motif: RKGR for the N-terminal and KRGK for the C-terminal domain, with both sequences conserved among species. To assess whether this could explain the different affinities for heparin, we designed two new ColQ mutants in which the BBXB motifs were exchanged between the sites. Having eliminated the Cterminal motif, the remaining N-terminal BBXB sequence was replaced by that of the C terminus and vice versa (Fig. 1B, Table I). As shown in Fig. 5B, the C3 N/C Ϫ mutant, which presents only the C-terminal BBXB motif located at the Nterminal position, eluted from the heparin-agarose column at the same NaCl concentration as N ϩ /C Ϫ . Similarly, N Ϫ /N3 C, which presents the N-terminal consensus sequence at the Cterminal position, behaved exactly as did N Ϫ /C ϩ . These results indicate that the different heparin affinities of the two HBDs are not explained by the identities of the basic residues that constitute each BBXB motif.
Contribution of Individual Basic Residues to Heparin Interactions-To evaluate individual residue contributions within the two binding sites, point mutations were carried out in each HBD to eliminate basic charges while trying to alter the local conformation as little as possible. Arginine or lysine residues located in the Xaa position as well as lysines in the Yaa position of Gly-Xaa-Yaa triplets were substituted by alanines, whereas proline was used to replace arginines in the Yaa position. Previous studies show arginine and proline in the Yaa position to be equally stabilizing, whereas the stability of arginine in the Xaa position as well of lysines in both positions were equivalent to alanine (22).
As shown in Fig. 6A, all N-terminal HBD mutants displayed decreased heparin affinities, suggesting the involvement of all residues from this HBD in heparin binding. According to the relative affinity loss presented by each mutant, the importance of each basic residue was ranked as Lys-122 Ͼ Arg-124 Ͼ Arg-118 Ն Arg-121.
In the C-terminal HBD, all mutations generated a significant change in heparin affinity (Fig. 6B). The mutation of any basic residue from the common GRPGBBGB sequence led to a decrease in heparin affinity. Residues were ranked in the following order of importance: Lys-226 Ͼ Lys-229Ͼ Ͼ Arg-227 Ͼ Ͼ Arg-223. Unexpectedly, the mutation of the most distal surrounding basic residues led to an increase in heparin affinity. The NaCl concentration required to elute N Ϫ /R218P did not differ significantly from the WT enzyme, in which both HBDs are intact, whereas that of N Ϫ /K233A surpassed it ( Table I).
Effect of Local Stability on Heparin Binding-From results shown in Fig. 5B, it is clear that neither the amount of basic charge nor the nature of the basic residues in the HBDs is the only determinant for heparin affinity, suggesting that structural factors such as local conformation could play a role. Proline is the most stabilizing residue found in collagens (21). Hence, basic residues from each HBD were selected to be replaced alternatively with alanine or proline. In the N-terminal HBD, substitution of the central lysine by proline caused a small affinity loss in contrast to the dramatic decrease observed for alanine substitution (Fig. 7A). In the C-terminal HBD, sub- Cumulative areas of the elution profiles from heparin-agarose of different A 12 AChE mutants. A, Lys-122, from the N-terminal HBD, was substituted either by alanine or by proline. In those mutants, the C-terminal HBD is inactivated. N ϩ /C Ϫ , which presents an intact Nterminal HBD, is shown for comparison. The elution value of K122P/C Ϫ was significantly higher than that of K122A/C Ϫ (p Ͻ 0.005, n ϭ 3). B, having an inactivated N-terminal HBD, Arg-218 or Arg-227 were substituted either by alanine or proline. In both cases the presence of a proline gave higher elution values than the presence of an alanine (p Ͻ 0.005, n ϭ 3 for Arg-218, and p Ͻ 0.000001, n ϭ 4 for Arg-227).
stitution of Arg-227 by proline induced a lower decrease in heparin affinity than substitution by alanine. For Arg-218, mutation to alanine decreased the affinity, whereas mutation to proline increased the affinity for heparin (Fig. 7B). In each case, affinity was higher with a proline than with an alanine.

Contribution of the Distant Environment to Heparin
Binding-Because local conformation seemed to be important in modulating heparin affinity, the possibility that regions distant from the HBDs could influence ColQ-heparin interactions was investigated. For this, ColQ deletions were designed to subject each HBD and its local environment to the same distant environment. The local environment was defined as the five triplets flanking each BBXB motif, so that each broader "mega"-HBD comprised 12 triplets. A single mega-HBD was then conserved, and the entire sequence (35 triplets) between the N-terminal limits of these domains was deleted (Figs. 1B and 8A). Thus, ⌬N contained the C-terminal HBD and vice versa for ⌬C, both surrounded by the most C-terminal and N-terminal regions of ColQ. Both constructs were able to produce truncated asymmetric AChE forms in Xenopus oocytes, exhibiting identical sedimentation profiles in sucrose gradients characterized by increased S values (Fig. 8B). As shown in Fig.  8C, ⌬N eluted from the heparin column at 0.5 M NaCl, a value lower than that of N Ϫ /C ϩ , whereas the ⌬C mutant eluted at 0.47 M NaCl, higher than N ϩ /C Ϫ (see Fig. 9 for a summary of relative heparin affinity of all the mutants). This indicates that although the C-terminal HBD still exhibits a higher affinity for heparin than the N-terminal HBD, these affinities converge when these sites are located in the same structural context, suggesting that even distant sequences in ColQ are able to affect the heparin affinity of each HBD.  ⌬N (right). Thus, the resulting molecules contain either the N-terminal HBD (⌬C) or the C-terminal HBD (⌬N), surrounded by their own immediate environment and having the same distant environments. The resulting A 12 AChE forms are also schematized. B, sedimentation profile on a sucrose gradient of AChE molecular forms expressed with ⌬C ColQ. The A 12 form (18 S) from the peak shaded in gray was isolated for heparin-binding assays. The same profile was obtained with ⌬N. C, elution profiles from heparin-agarose columns, expressed as cumulative areas, from AChE mutants presenting only one active HBD. In N ϩ /C Ϫ and in ⌬C, only the N-terminal HBD is active, whereas in N Ϫ /C ϩ and ⌬N, the C-terminal HBD is still intact. When comparing the elution values, ⌬N had more affinity for heparin than N Ϫ /C ϩ (p Ͻ 0.005, n ϭ 3), and ⌬C less affinity than N ϩ /C Ϫ (p Ͻ 0.001, n ϭ 3), but still ⌬N presented significantly higher heparin affinity than ⌬C (p Ͻ 0.005).
FIG. 9. Diagrammatic representation of the relative affinities for heparin of asymmetric AChE mutants. Relative affinity for heparin is expressed as NaCl concentration required to elute 50% of the enzyme bound to a heparin-agarose column. The different mutants are grouped in four columns. Left to right, mutants lacking both HBDs, mutants having only the N-terminal HBD, mutants having only the C-terminal HBD, and mutants having both HBDs. As a reference, the WT enzyme together with N ϩ /C Ϫ (has only the N-terminal HBD), N Ϫ /C ϩ (has only the C-terminal HBD), and N Ϫ /C Ϫ (both HBDs are inactivated) are highlighted in gray.
ColQ confers on it the capacity to interact with multiple ligands, which may allow precise regulation of its localization. Several lines of evidence have suggested that ColQ-HSPG interactions are responsible for retention of AChE in the basal lamina and for its specific localization at the neuromuscular junction. In agreement with this idea, the collagen domain of ColQ contains two basic clusters that we have proposed to mediate ColQ-HSPG interactions by binding to HS chains. We have previously characterized those putative HBDs by molecular modeling and by the use of synthetic peptides. In this work, by analyzing the entire recombinant asymmetric AChE, we show that (i) the heparin-binding capacity of ColQ resides exclusively in two HBDs, (ii) the two HBDs exhibit different properties, and (iii) heparin affinity is determined by the combined contribution of sequence and triple-helix conformation. Besides its relevance to AChE localization, this study contributes to the general understanding of collagen-ligand interactions.
Heparin-binding Capacity of ColQ Resides Exclusively in the Two HBDs Present in Its Collagen Domain-In addition to the HBDs in the collagen domain of ColQ, its C-terminal domain has also been implicated in asymmetric AChE localization at the neuromuscular junction. In congenital myasthenic syndrome type 1c patients, several mutations in the ColQ gene impair folding of ColQ, avoiding enzyme secretion. In some cases, there are point mutations in the C-terminal domain of ColQ in which asymmetric AChE forms are produced but are not accumulated at the neuromuscular junction (2). Because the interaction of AChE with HSPGs has been proposed to represent a crucial step for AChE retention and localization, it has opened the question of whether these mutations in the C-terminal domain of ColQ modify its interaction with HSPGs. We assessed this possibility and unambiguously established that only the collagen domain of ColQ interacts with heparin. AChE oligomers containing the collagen domain with or without a C-terminal domain presented a full binding to heparin, whereas oligomers lacking the collagen domain lost all binding capacity.
Heparin Interaction Depends upon Basic Residues, but Special Rules Apply in Collagens-Heparin binding has been extensively studied in globular proteins. In most cases, basic residues interact via electrostatic interactions with sulfate or carboxylate groups in the heparin molecule (23). However, in certain proteins such as thrombospondin, heparin binding involves tryptophan and serine residues (24). For ColQ, we showed that the interaction is primarily electrostatic, since the elimination of basic charges completely abolishes heparin binding. This is consistent with chemical modification studies, in which maleylation of the ⑀-NH 2 groups of lysine residues eliminated the tendency of asymmetric AChE to form glycosaminoglycan-dependent aggregates (25), and with calorimetric measurements that indicate that heparin binding is predominantly electrostatic. 2 For heparin-binding proteins where basic residues are involved, different consensus sequences have been described according to the structural context in which they are expressed. Whereas the BBXB sequence is found preferentially in ␤-strands, the most frequent sequence in ␣-helices is BBBXXB, generating in both cases a segregation of the basic residues on one face of the structure (26). Although these motifs are absent in several heparin-binding collagens (27), one such consensus sequence was present in ColQ and helped us to initially define the putative HBDs (11). Sequences that interact with heparin have been identified in collagens type I and type V/XI. In type I collagen ␣1(I) 2 ␣2(I), the residues that have been implicated in heparin binding are concentrated only in two triplets (i.e. KGHRG in ␣1(I) and KGIRGH in ␣2(I)) (28), whereas in collagens ␣1(V) 3 and ␣1(XI)␣2(XI)␣1(II) basic residues are spread throughout the sequence (i.e. KXGPRGXRGPTGPRGXR present in ␣1(V), ␣1(XI), and ␣2(XI) chains) (29). In ColQ, our results showed that the most important residues for heparin interaction are concentrated in the sequence GRPGBBGB, differing from those found in collagens. They all have in common the presence of basic residues, but no consensus sequence for triple-helical proteins seems to emerge.
The nature of basic residues is an important factor for heparin binding. Arginine binds tighter to heparin than lysine, since it has a higher potential for forming hydrogen bonds, and the guanidinium group may form an intrinsically stronger electrostatic interaction with a sulfate anion than the ammonium group (30). This has been clearly shown using non-collagenous heparin-binding peptides containing different proportions of arginine and lysine residues (30,31). However, in each HBD we found the most important residue to be a lysine. This could be a particularity of collagen, where arginines in Yaa position play an important role in the stability of the triple-helical structure (22). Molecular modeling studies show that arginines can form hydrogen bonds with the backbone of the same or the neighboring polypeptide chains (32). This was frequently observed in the model of ColQ, where arginine side chains were packed against the triple helix (12), reducing their availability to interact with other molecules. This seems to be the case of two arginines in Yaa position located C-terminal to the GRPBBGB motif in the N-terminal HBD of Torpedo that are not conserved in mammals, where these positions are occupied by prolines (Fig. 1A). Consistent with this, a decrease in stability was observed when these arginines were replaced by alanines in triple-helical peptides modeling the Torpedo N-terminal HBD (13). Thus, depending on the contacts they establish, arginines could alternatively participate directly in heparin interactions or play a structural role, which in turn can affect the accessibility of the neighboring residues to the ligand. In collagens, interaction with heparin involves basic residues coming from different polypeptide chains (12), and the contribution of basic residues varies depending on the collagen chain composition. Thus, it seems very difficult to define a general consensus sequence for heparin binding in triple-helical collagens.
Triple-helix Structural Properties as Modulators of Heparin Affinity-Previous studies using synthetic peptides have shown a negative correlation between affinity and triple-helical stability (13). Here, ColQ mutants in which a same residue was substituted by alanine or proline, generating an identical array of basic residues but likely to be subjected to different local stability, showed different affinities for heparin. However, contrary to the peptides, we found that the presence of a proline, a residue that stabilizes triple-helical conformation, generated a mutant enzyme with more affinity for heparin than the presence of an alanine. A possible explanation for this discrepancy is that the 30-mer peptides and the full-length ColQ present different levels of local stability. In fact, in order to force the peptides to adopt a triple-helical conformation in solution, it was necessary to surround the HBD sequences with Gly-Pro-Hyp triplets, the most stabilizing sequences in collagen (21). On the other hand, the HBDs in ColQ are surrounded by triplets with low proline content, and as shown by molecular modeling, this can lead to a local untwisting of the triple helix facilitated by the repulsion between chains created by the high basic charge content (12). We propose that affinity for heparin is optimal in an intermediate range of local stability. In highly stable regions the triple helix is too rigid to be able to accommodate the ligand, whereas in very low stability regions the three polypeptide chains could be locally untwisted so that the geometrical arrangement of basic residues necessary for heparin binding would not be generated.
Local conformation, and hence heparin affinity, seems also to be influenced by the distant sequences, as seen in the ⌬N and ⌬C mutants. When surrounded by the same distant environment, the two HBDs showed closer affinities for heparin than when located in their original individual environment. A component of the distant environment would be the structural features of the surrounding triple-helical sequences per se, and these may in turn be influenced by the constraints imposed by the other domains in ColQ, as by the presence of 12 catalytic subunits. The effect of the distant environment may also explain the fact that the wild type enzyme with both HBDs exhibits a greater affinity for heparin than each HBD independently. We have previously shown that triple-helical content and stability increase upon heparin binding to a HBD (14). Thus, such a perturbation induced by heparin binding to one HBD could propagate along the triple helix, altering the conformation of the other HBD and increasing its affinity for heparin. Considering that collagen thermal stability is so close to body temperature, it is possible to speculate that its conformation will be sensitive to subtle changes in the milieu, allowing for a fine-tuned regulation of the interaction with other molecules.
Two Different HBDs: More Than One Function?-The presence of two HBDs with different properties that are conserved between Torpedo and mammals opens new perspectives regarding their function in AChE localization and, more generally, in basal lamina organization.
A possible interpretation for the differences observed between the two HBDs is that they reflect differences in specificity. This specificity can operate at different levels. One possibility is that each HBD has a preference for specific microdomains in the HS molecule, determined by specific sulfation patterns (33). The fact that distal basic residues in the C-terminal HBD seem to interfere with heparin binding supports this idea. This HBD, with an ample distribution of basic charges, is most likely a binding site for HS, whereas the GRPGBBGB motif with more concentrated basic charges would be preferred by heparin-like (i.e. highly sulfated) domains in HS. These specific sulfation microdomains could be located in different regions of a single HS chain, in different HS chains of a single proteoglycan, or even in different proteoglycans. Another possibility is that each HBD presents a specific preference for different glycosaminoglycans. In this context, a possible candidate is chondroitin sulfate. It has been suggested previously that chondroitin sulfate proteoglycans interact with asymmetric AChE (34,35), and we have observed previously a direct interaction between chondroitin sulfate and the peptides modeling the N-and C-terminal HBDs, whose characteristics also differed between the two sites (14).
We propose that the glycosaminoglycan microdomains recognized by each HBD would be located in different layers of the basal lamina. It is interesting to note that the length of ColQ triple helix, measured both in electron micrographs (3,25) as in the structural model (12), is around 50 nm, which corresponds to the distance between pre-and postsynaptic membranes (36). Therefore, the distance between the two HBDs (ϳ25 nm) represents an important fraction of the space between these mem-branes, so that their binding to different layers of the extracellular matrix would serve to orient the enzyme for proper function.
Despite its "simple" and linear structure, the collagen triple helix provides a very complex scenario in terms of ligand interactions, in particular heparin interactions. The importance of the local conformation of the triple helix as a key modulator of the interaction with heparin, and potentially with HSPGs, opens new perspectives on the dynamic of the basal lamina.