Substrate and Product Trafficking through the Active Center Gorge of Acetylcholinesterase Analyzed by Crystallography and Equilibrium Binding*

Hydrolysis of acetylcholine catalyzed by acetylcholinesterase (AChE), one of the most efficient enzymes in nature, occurs at the base of a deep and narrow active center gorge. At the entrance of the gorge, the peripheral anionic site provides a binding locus for allosteric ligands, including substrates. To date, no structural information on substrate entry to the active center from the peripheral site of AChE or its subsequent egress has been reported. Complementary crystal structures of mouse AChE and an inactive mouse AChE mutant with a substituted catalytic serine (S203A), in various complexes with four substrates (acetylcholine, acetylthiocholine, succinyldicholine, and butyrylthiocholine), two non-hydrolyzable substrate analogues (m-(N,N,N-trimethylammonio)-trifluoroacetophenone and 4-ketoamyltrimethylammonium), and one reaction product (choline) were solved in the 2.05-2.65-Å resolution range. These structures, supported by binding and inhibition data obtained on the same complexes, reveal the successive positions and orientations of the substrates bound to the peripheral site and proceeding within the gorge toward the active site, the conformations of the presumed transition state for acylation and the acyl-enzyme intermediate, and the positions and orientations of the dissociating and egressing products. Moreover, the structures of the AChE mutant in complexes with acetylthiocholine and succinyldicholine reveal additional substrate binding sites on the enzyme surface, distal to the gorge entry. Hence, we provide a comprehensive set of structural snapshots of the steps leading to the intermediates of catalysis and the potential regulation by substrate binding to various allosteric sites at the enzyme surface.

strates, including ACh, inhibit their own catalysis at concentrations exceeding the K m (Ն1 mM, i.e. Ն20 ϫ K m for ATCh and mouse AChE (mAChE) (15)). AChE from Drosophila, but not from vertebrates, also shows substrate activation at low concentrations (23,24). Competitive displacement, by ACh, of the specific PAS ligands, propidium and the peptidic toxin fasciculin, provides evidence that ACh binds the PAS in addition to the active site (10,25,26).
Electrostatic calculations based on the TcAChE structure and subsequent molecular dynamics simulation suggested that AChE may also have a "back door," distinct from the gorge entrance and whose transient opening would contribute to the high rate of traffic of substrates, products, and water into and out of the active center gorge (27,28). Opening of this putative door would involve a displacement of the Trp 86 side chain, which constitutes the thin wall separating the choline binding site in the active site from the outside solvent. Existence of residual catalytic activity of AChE complexes with the large fasciculin molecule, which binds the PAS to seal the gorge entrance (5,18), may also argue for the need of an alternative entry portal(s) for the substrate (26, 29 -31). Yet, crystal structures of fasciculin-AChE complexes did not reveal an open back door. To date, the functioning of the back door remains hypothetical and, apart from some speculation (32,33), evidence for its existence has not been reported.
Although a structural perspective on the AChE catalytic mechanism should include the acylation and deacylation steps and related intermediates, the rapid substrate turnover precludes their entrapment and visualization in a crystalline state. Initial manual docking of an ACh molecule into the Torpedo californica (TcAChE) structure (4) was followed by structures of AChE complexes with competitive, reversible inhibitors (34 -38), covalent organophosphate or carbamate inhibitors (39), bifunctional inhibitors (20,32,34,40,41), PAS inhibitors (5,18,19), and the substrate analogue, trimethylammoniotrifluoroacetophenone (TMTFA) (42). The latter provides a close structural mimic of the substrate-AChE tetrahedral transition state.
To enhance the opportunity of observing intact substrate molecules bound with high occupancies at the various sites, we expressed a catalytically inactive mAChE mutant, S203A, and crystallized it together with the wild type mAChE in a form that keeps potential ligand binding sites at the enzyme surface free of packing contacts (19,20). We then soaked the crystals in concentrated solutions of the substrates ACh and acetylthiocholine (ATCh), the slowly hydrolyzable substrates succinyldicholine (SCh) and butyrylthiocholine (BTCh), the non-hydrolyzable substrate analogues TMTFA and 4-ketoamyltrimethylammonium (4K-TMA), and the reaction product choline (Scheme 1). The eight resulting crystalline complexes and conjugates, supported by binding and inhibition data, show complementary views of the transition state, the acyl-enzyme and the deacetylated enzyme, and substrate binding to the PAS. In addition, substrate binding sites remote from the active site and the PAS were identified. Hence these structures provide a comprehensive set of snapshots of the reaction intermediates along the gorge pathway and the likely modes of regulation by substrate binding to low affinity sites at the enzyme surface.

MATERIALS AND METHODS
Chemicals-4K-TMA iodide was from ICN and ambenonium dichloride from Tocris Cookson. ACh, ATCh, SCh, BTCh, and choline, as chloride or iodine salts, were from Sigma. TMTFA iodide and decidium diiodide were respective gifts from Daniel M. Quinn, University of Iowa, Iowa City, UT, and Harvey A. Berman, State University of New York, Buffalo, NY. Polyethylene glycols were from Hampton Research, Fluka, or Sigma.
Enzymes-Soluble mAChE expressed in human embryonic kidney-293 cells (15,43) was purified by affinity chromatography with elution using either 5 mM propidium diiodide (19) or 100 mM SCh dichloride; in the latter case, the eluted enzyme was immediately dialyzed to avoid precipitation/inactivation due to medium acidification through slow SCh hydrolysis. The mAChE mutant S203A was generated by site-directed mutagenesis, expressed in human embryonic kidney-293 cells, and purified using 100 mM decamethonium bromide (15,43). The enzyme and mutant were extensively dialyzed against 1 mM Mes, pH 6.5, 1 M NaCl, 40 mM MgCl 2 , 0.01% NaN 3 (w/v), and desalted by gel-filtration FPLC on Superdex-200 (Amersham Biosciences) in 1 mM Mes, pH 6.5, 50 mM NaCl, 0.01% NaN 3 (w/v) (crystallization buffer), or by extensive dialysis against this buffer. They were concentrated to about 10 mg/ml by ultrafiltration.
Crystallization, Complex Formation, and Data Collection-Crystallization was achieved at 4°C by vapor diffusion using hanging drops (1-2 l) and a protein-to-well solution ratio of 1:1, with P550MME or P600 (25-32%) (v/v) in either Hepes or sodium acetate (60 -100 mM), pH 6.5-8.0, as the well solution (19). The TMTFA-mAChE complex was formed in solution using apo-mAChE at 5.0 mg/ml (ϳ80 M; 1.5 10 10 ϫ K i (31)) and a 3-fold molar excess of the ligand over the enzyme, and concentrated prior to crystallization. The other seven complexes were generated by crystal soaking, carried out at 4°C in sitting drops (20 l) made of the well solution supplemented with the ligand (250 mM) and polyethylene glycol up to 35% SCHEME 1. The ligands used in this study. Shown are the substrates ACh, ATCh, SCh, and BTCh, the substrate analogues TMTFA and 4K-TMA, and the reaction product choline.
(ϳ0.6 M) to ensure cryoprotection; the soaking drops were renewed twice at 12-h intervals and again just before direct crystal flash-cooling in the nitrogen gas stream. Crystals belong to the orthorhombic space group P2 1 2 1 2 1 with unit cell dimensions a ϭ 79 Å, b ϭ 112 Å, c ϭ 227 Å. Oscillation images were integrated with DENZO (44) and data were scaled and merged with SCALA (45) ( Table 1).
Structure Determination and Refinement-The apo-mAChE structure (Protein Data Bank entry 1J06 (19)) without solvent was used as a starting model to refine the reported structures with REFMAC (46) using the maximum likelihood approach and incorporating bulk solvent corrections, anisotropic F obs versus F calc scaling and TLS refinement with each subunit defining a TLS group (Table 1). Typically, rigid-body refinement was first performed on each of the two subunits forming the crystalline mAChE dimer (19) using all data followed by cycles of restrained refinements. Random sets of reflections were set aside for cross-validation purposes. For each structure, the resulting A-weighted 2F o Ϫ F c and F o Ϫ F c electron density maps were used to position the ligand. Automated solvent building was performed with ARP/wARP (47) and manual adjustment with the graphics program TURBO-FRODO (66) and COOT (48).
The final structures, one apo-S203A and eight S203A or mAChE complexes, comprise residues Glu 1 -Thr 543 and Glu 4 -Thr 543 for the two mAChE/S203A molecules in the asymmetric unit, respectively, and GlcNAc moieties linked to Asn 350 and Asn 464 . High temperature factors and weak electron densities are observed for residues within the sur-face loop region Asp 491 -Pro 498 . The average root mean square (r.m.s.) deviation value between the nine structures and the apo-mAChE structure is 0.24 Å for 535 C␣ atoms (from 0.19 Å for the SCh-mAChE complex to 0.294 Å for the 4K-TMA-mAChE complex). The stereochemistries of the structures were analyzed with PROCHECK (49); with the exception of residue Ser/Ala 203 , no residues were found in the disallowed regions of the Ramachandran plot. The atomic coordinates and structure factors of the four S203A (ACh, ATCh, SCh, and BTCh) and four mAChE complexes (TMTFA, 4K-TMA, SCh, and choline) have been deposited with the RCSB Protein Data Bank (Table 1). Figures were generated with PyMOL (50).
Inhibition and Binding Studies-Inhibition by SCh, 4K-TMA, and choline of mAChE-catalyzed ATCh hydrolysis was measured spectrophotometrically (51) in 0.1 M phosphate buffer, pH 7.0, at 22°C. The inhibition constants, K i and ␣K i , were determined from the dependence of slopes and y intercepts of Lineweaver-Burk plots on inhibitor concentration (15).
Binding of ligands ACh, ATCh, SCh, BTCh, 4K-TMA, and choline to mAChE (200 nM) and the S203A mutant (100 nM) was measured from the dependence of the pseudo-first order association rate, k obs , of the reversible bisquaternary inhibitors ambenonium (2 or 5 M) (52) or decidium (1 M) (53) on ligand concentration (from 1 M to 30 mM for 4K-TMA and to 300 mM for the other ligands). Rates were monitored in a millisecond time frame by stopped-flow measurements of intrinsic Trp fluorescence quenching of mAChE or its mutant as where I is an individual reflection measurement and ͗I ͘ is the mean intensity for symmetry-related reflections.
where F o and F c are observed and calculated structure factors, respectively. R free is calculated for 2% of randomly selected reflections excluded from refinement e Root mean square deviations from ideal values. described (54). Dissociation constants, K d , were determined from semi-logarithmic plots of k obs versus ligand concentration ([L]) and calculated by nonlinear regression using Equation 1, where k obs 0 is the pseudo-first order rate constant for ambenonium or decidium association in the absence of ligand. When k obs versus [L] plots were biphasic, two dissociation constants, K d and K d L , characterizing ligand binding to the enzyme and the enzyme-ligand complex, respectively, were calculated using Equation 2, where k obs L,0 is the limiting first-order rate constant for competitor association with the enzyme-ligand complex.

Procedures for Trapping Intact Substrate in the Crystalline Enzyme
The 2.65-Å resolution structure of the S203A mutant, compared with apo-mAChE (r.m.s. deviation value: 0.26 Å for 535 C␣ atoms), shows an undisrupted catalytic site architecture despite the absence of the Ser hydroxyl, as well as unaltered topography for the PAS region ( Fig. 1). Soaking a S203A crystal with 50 mM ATCh (ϳ10 3 ϫ K m and 4 ϫ K ss for mAChE; ϳ50 ϫ K d for the mutant; Table 2) led to a 2.25-Å resolution structure in which only the reaction products thiocholine (TCh) and acetate were found, well ordered and with high occupancy, in the active center gorge (not shown). Despite the small size of choline, the side chain of Tyr 337 , located in the constricted region of the gorge, was found rotated by 24°and tilted by 35°compared with its position in apo-mAChE, a movement that increases the gorge internal diameter and may favor choline release in catalysis. Subsequent analyses of a choline-S203A and a choline-mAChE complex, both formed with 50 mM choline (15-50 ϫ K d ; Table 2), showed bound choline molecules in the same position and orientation as the bound TCh (not shown). This not only indicates that the thioester bond in the substrate trapped in the S203A crystal was cleaved by the synchrotron radiation during data collection, but also that absence of the Ser 203 hydroxyl does not alter orientation of the choline moiety, despite the 4 -10-fold lower affinities of the mutant for  ATCh and choline (Table 2). To increase the probability of observing intact substrate molecules bound with high occupancies, we employed concentrations greater than the substrate inhibition constants (250 mM, i.e. ϳ20 ϫ K ss for ACh and mAChE; higher concentrations led to ill-diffracting crystals), and substrates or analogues with oxo or keto bonds expected to be less sensitive to radiation (Scheme 1).

Determination and Quality of the Structures
The structures of mAChE in complexes with the ACh keto analogues, TMTFA and the isosteric 4K-TMA, the poor substrate SCh, and the product choline, and those of the S203A mutant in complexes with substrates ACh and ATCh and poor substrates SCh and BTCh, were solved by difference Fourier methods and refined in the 2.05-2.65-Å resolution range, with good R-factor values and stereochemistries (Table  1). For all eight complexes, the excellent quality of the final electron density maps permits an unambiguous positioning of the ligands, which are well ordered and bound with high occupancy. The polypeptidic chain, which folds into a 12-stranded centralmixed ␤-sheet surrounded by 14 ␣-helices, adopts the same conformation as found in the apo-mAChE and S203A structures (r.m.s. deviation values in the 0.19 -0.29 Å range for 534 C␣ atoms). This indicates that major conformational changes are not associated with ligand binding. Only a rotation, up to 26°, and a tilt, up to 50°, of the Tyr 337 side chain is observed throughout, which may simply reflect the high mobility of this residue in the crystaline state (19,20,55). Hence, these structures, when complemented by kinetic analyses performed on the same complexes, provide a comprehensive set of snapshots for exploring the routes for trafficking of substrate into and product of the active center gorge, as well as the location of alternative substrate binding sites at the enzyme surface.

Substrate and Product Binding in the Active Site and at the Peripheral Anionic Site
TMTFA-mAChE Complex-The structure of the co-crystallized TMTFA-mAChE complex shows the substrate analogue oriented perpendicular to the gorge axis and positioned within covalent bond distance of the Ser 203 hydroxyl in the active site at the base of the gorge (Fig. 2A). The trimethylammonio group is ideally positioned for cationinteraction with the 6-membered ring of Trp 86 , located 4.2 Å away. The carbonyl, which points into the oxyanion hole, is hydrogen bonded to Ala 204 and the Gly 121/122 backbone amines, whereas the trifluoromethyl group is anchored within the acyl pocket, bordered by Trp 236 , Phe 295 , Phe 297 , and Phe 338 . This position and orientation of the bound TMTFA are consistent with those found in the TMTFA-TcAChE complex (42) (r.m.s. deviation value: 0.8 Å for 518 C␣ atoms). A well defined polyethylene glycol molecule (arising from the crystallization liquor) is bound at the PAS in a similar position as in apo-mAChE (19). Hence, this structure provides a frame of reference for specific versus nonspecific ligand binding at the PAS, both altering the shape of the electron density maps.
4K-TMA-mAChE Complex-In the 4K-TMA-mAChE complex, two well ordered molecules of the substrate analogue are, respectively, bound in the active site, and at the PAS at the gorge entry (Fig. 2B). The 4K-TMA bound in the active site, oriented as is TMTFA bound at the same site (cf. Fig. 2A), adopts a near perfect non-relaxed trans, trans conformation. The carbonyl carbon is within covalent bond distance (1.39 Å) of the Ser 203 hydroxyl, whereas the trimethylammonio group faces (4.2 Å) the six-membered ring of the Trp 86 indole. The position and conformation of 4K-TMA in the mAChE active site resembles those of the ACh molecule manually docked into the active site of TcAChE (4), but the quaternary (cationic alkyl) ammonium group deviates by 1 Å from its position in docked ACh as to face the five-membered ring of the Trp 286 indole (Fig. 3A). This distinctive position also coincides with that of the corresponding atoms of TMTFA bound to mAChE (cf. Fig.  2A). The tetrahedral conformation of the 4K-TMA carbonyl carbon, along with its short distance to the Ser 203 hydroxyl, indicates that a perfect tetrahedral hemiketal conjugate was generated and trapped in the crystalline enzyme.
The 4K-TMA molecule bound at the PAS surface aligns along the gorge axis and adopts the same nonrelaxed trans, trans conformation as in the active site, except for maintaining a trigonal, planar instead of tetrahedral configuration (Fig. 2B). The quaternary ammonium group establishes cationinteraction (4.4 Å) with the Trp 286 indole and van der Waals contacts (4 Å) with Tyr 72 at the gorge entrance. The keto moiety, positioned between Tyr 124 and Tyr 341 deeper in the gorge, has its carbonyl oxygen linked to the Phe 295 backbone nitrogen atom via a 2.6-Å distant iodide ion (the added counter ion to 4K-TMA), and with the residue pair Asp 74 /Tyr 341 on one side and its methyl in van der Waals contacts with Tyr 124 on the other side. Hence the PAS-bound 4K-TMA is inverted with respect to the dipole moment of the gorge.
Comparison of the 4K-TMA-mAChE complex with apo-mAChE (r.m.s. deviation value: 0.27 Å for 533 C␣ atoms) shows limited conformational changes associated with 4K-TMA binding (not shown). Some side chains reorient, such as those of Tyr 341 and Asp 74 at the PAS, which move slightly into the gorge, and of Tyr 337 , which rotates by ϳ8°to enlarge the gorge diameter near its constricted region, along with small movements of Trp 86 (ϳ4°) in the choline binding site and Trp 286 (ϳ17°) at the PAS. A subtle rotation, by 18°, of the Ser 203 hydroxyl helps position the tetrahedral conjugate to optimize occupation of the oxyanion hole.
Kinetics studies show that 4K-TMA is a reversible, nonhydrolyzable substrate analogue with lower affinity for the S203A mutant than for mAChE ( Table 2). In fact, soaking of 4K-TMA in the S203A crystal led to a structure devoid of a bound ligand (not shown), consistent with the role of Ser 203 for stabilizing the analogue as a tetrahedral hemiketal within the intact active center of mAChE. However, the basis for the absence of bound 4K-TMA at the PAS of the mutant is unclear.
Acetylcholine-and Acetylthiocholine-S203A Complexes-In the ACh-S203A complex, two well ordered ACh molecules are found in the active site and at the PAS, respectively (Fig. 2C). The ACh positioned in the active site adopts a relaxed trans, gauche conformation, consistent with the most stable conformation of ACh seen by crystallography and NMR spectroscopy (56,57). This bound ACh has its quaternary ammonium group facing Glu 202 (4.1 Å) and in a cation-interaction with the 6-membered ring of Trp 86 (4.3 Å), whereas the acetoxy moiety, located near the constricted region of the gorge, interacts via the oxygen with Tyr 124 on one side of the gorge and, via the methyl, with Tyr 337 and Phe 338 on the other side. An acetate molecule (arising from the crystallization liquor) occupies the oxyanion hole, as do carbonate or acetate molecules found in previous mAChE structures (19).
ACh bound at the PAS adopts a position and a non-relaxed trans, trans conformation similar to those of 4K-TMA bound at the same site (Fig. 2C, and cf. Fig. 2B). As a result, the PASbound ACh is inverted with respect to the molecule bound in the active site, such that the two acetyl groups are separated by only 4.5 Å in the constricted region of the gorge. These two ACh orientations at the PAS and at the base of the gorge may reflect two successive snapshots of ACh trafficking into the gorge. The orientation of the PAS-bound substrate may also limit its entry to the active center, when inhibition of catalysis by excess substrate occurs. The largest differences compared with apo-mAChE are confined to a ϳ20°rotation of the Tyr 337 phenol that increases the gorge diameter at mid-point and a ϳ10°rotation of Trp 286 that enlarges the gorge entrance.
In the ATCh-S203A complex (Fig. 2D), near equal amounts of intact ATCh, in a relaxed trans, gauche conformation as seen in the ACh-S203A complex (cf. Fig. 2C), and of TCh (arising from partial cleavage of the thioester bond during data collection) are bound in the catalytic site, whereas an ethylene glycol molecule (arising from the crystallization liquor) occupies the oxyanion hole. A second TCh molecule (presumably, also generated in the active center during data collection) is bound at the PAS. In the active site, the overlapped choline moieties in the substrate and product face Glu 202 (4.2 Å) and are in a cation-interaction with the nearly parallel Trp 86 indole. The position of the free sulfhydryl in TCh creates unfavorable interactions with the Tyr 337 hydroxyl (2.1 Å) and Trp 286 indole (3.2 Å). Compared with apo-mAChE, a 25°rotational movement of Tyr 337 enlarges the gorge in the region of constriction (not shown). The presence of a non-attributed electron density with peaks above 4.5 near the Tyr 337 phenol ring and negative peaks for the tip of this ring indicate even higher mobility of Tyr 337 in this complex than in the others. The overlay of ATCh-bound S203A with BTCh-bound somaninhibited butyrylcholinesterase (BChE) (58) (r.m.s. deviation value: 0.87 Å for 500 C␣ atoms) reveals a common substrate orientation dictated by the position of the quaternary ammonium groups, anchored near Trp 86 , whereas the two acyl chains diverge by 90°due to the Y337A substitution in BChE (Fig. 3C). At the PAS, the bound TCh, inverted with respect to the PASbound 4K-TMA in mAChE and ACh in the mutant (cf. Fig. 2, B and C), establishes similar cation-interactions associated with side chain repositioning as in the other complexes.
Succinylcholine-S203A and -mAChE Complexes-The structures of the S203A mutant and mAChE in complexes with the poor, slowly deacylating substrate SCh (Table 2) are similar to each other (Fig. 2, E and F). Both show a symmetrical SCh molecule that spans the full-length of the gorge with its two choline moieties, respectively, bound into the anionic subsite of the active center and at the gorge entrance. These moieties resemble, in their positions and conformations, those of the two inverted ACh molecules bound to the S203A mutant (cf. Fig.   2C). SCh binding is associated with a loop-forming bend of the succinate moiety, resulting in a S-shaped conformation for SCh. The two central succinyl carbonyls are tilted by ϳ127°c ompared with 180°in an extended trans conformation. This conformation facilitates cation-interactions of the two distant SCh quaternary ammonium groups with the Trp 286 and Trp 86 indole rings, respectively, in the PAS and the choline binding subsite of the active center. The binding interactions are dominated by van der Waals contacts with the apolar side chains that line the gorge (Phe 297 , Tyr 337 , Phe 338 , and Tyr 341 ), whereas a few polar interactions, mediated by solvent molecules, involve one carbonyl oxygen of the succinyl moiety and the Tyr 124 hydroxyl and Asp 74 carboxylate. The major conformational differences compared with apo-mAChE involve the side chain of Tyr 337 , which rotates by 26°and is tilted by 68°, and that of Trp 286 that moves by 9°. The drastic rotation of Tyr 337 promoted by the bound succinyl moiety induces a significant increase, by ϳ2.3 Å, of the gorge diameter. These displacements suggest that similar enlargements occur in solution facilitating substrate/product trafficking within the gorge.
In the SCh-mAChE complex, but not the SCh-S203A complex, a bifurcated electron density deviating from that accounting for the deepest SCh ether oxygen is clearly visible near the catalytic site (Fig. 2F). Perfect filling of this density by a succinylmonocholine molecule anchored to the catalytic Ser 203 , and oriented as are TMTFA and 4K-TMA bound to mAChE (cf. Fig.  2, A and B), reveals that an succinyl-enzyme intermediate (distance of 1.6 Å from the Ser hydroxyl to the succinylmonocholine carbonyl carbon) was trapped in the crystal, in near-equal amounts as intact SCh and with overlapping choline moieties proximal to Trp 86 . Entrapment of an acyl-enzyme intermediate in crystalline mAChE suggests that, compared with ACh, the slower rate of SCh hydrolysis ( Table 2) may reflect a slower rate of acylation and/or deacylation. In the SCh-S203A complex, an acetate molecule (arising from the crystallization liquor) is bound within the oxyanion hole in a position that roughly coincides with that of the acyl moiety attached to Ser 203 in the SCh-mAChE complex (Fig. 2, E and F).
Butyrylthiocholine-S203A Complex-Hydrolysis of BTCh catalyzed by mAChE is about 100 times slower than that of ACh or ATCh despite only 4-fold lower K d and 2-fold lower K m values ( Table 2). The structure of the S203A mutant in complex with BTCh shows a butyrate and a TCh molecule (both arising from cleavage of the thioester bond during data collection) bound in the catalytic site and an intact BTCh molecule bound at the PAS (Fig. 2G). In the catalytic site, TCh adopts the same position as in the ATCh-S203A complex (cf. Fig. 2D), whereas the butyrate occupies the oxyanion hole, as do acetate and ethylene glycol in other structures (cf. Fig. 2, C-E). Overlay of this structure with that of butyrate-bound BChE (58) (r.m.s. deviation value: 0.9 Å for 503 C␣ atoms) shows the butyryl chain oriented perpendicular to its position in BuChE due to differences in the shape of the acyl pocket between the two enzymes.
The BTCh molecule bound at the PAS adopts a position and a non-relaxed trans, trans conformation similar to those of 4K-TMA, ACh, and ATCh bound at the same site (cf. Fig. 2,  B-D, and G). The quaternary ammonium is oriented toward the gorge entrance. The butyryl moiety, anchored 2.8 Å deeper in the gorge than the acetyl moiety of ATCh (cf. Fig. 2D), is in the stacking interaction with Tyr 341 , whereas its carbonyl oxygen is linked to the Phe 295 backbone amide nitrogen via a 2.75-Å distant iodide ion (the added counter ion to BTCh), similarly as found in the 4K-TMA-mAChE complex. Again, a 17°rotation of Tyr 337 enlarges the diameter in the constricted region of the gorge.
The presence of intact BTCh at the PAS, but not in the active center of the S203A mutant, suggests that the lower affinity of the mutant for BTCh is entirely due to absence of the Ser 203 hydroxyl ( Table 2). This also suggests that the trans, gauche conformation of the substrates in the active center (cf. Fig. 2, C and D) is more sensitive to radiation than the trans, trans conformation adopted at the PAS (cf. Fig. 2, B, C, and G).
Choline-mAChE Complex-The structure of mAChE bound with the reaction product shows two choline molecules, respec-tively, bound near the catalytic site and at the PAS (Fig. 2H). At the active center, the choline quaternary ammonium group faces Glu 202 (3.9 Å) and establishes a cationinteraction (4.3 Å) with the sixmembered ring of Trp 86 , with the three methyls oriented nearly parallel to the indole ring. The primary alcohol carbon is in van der Waals contact with the Tyr 337 phenol, which is oriented perpendicular to Trp 86 , whereas the oxygen is bound to the His 447 backbone carbonyl oxygen. In the oxyanion hole, a water molecule is bound to the Ser 203 hydroxyl and occupies the same position as the acetate, ethylene glycol, or butyrate found in other structures (cf. Fig. 2, C-E and G). The overlay of the choline-mAChE and choline-BChE (58) complexes (r.m.s. deviation value: 0.95 Å for 501 C␣ atoms) shows similarly oriented bound choline molecules with their quaternary ammonium groups separated by only 0.25 Å near the acyl pocket and overlapping primary alcohols directed toward the gorge entrance (Fig. 3B). This orientation of the bound choline, compared with bound ACh or ATCh in the S203A mutant (cf. Fig. 2, C and D), argues for a ϳ90°flip of the choline product upon dissociation before it leaves the gorge.
At the PAS, the choline quaternary ammonium group establishes cationinteraction with Trp 286 (4.4 Å) and van der Waals contact with Tyr 124 and Tyr 341 , whereas the oxygen atom points outside the gorge and establishes a weak (long range) hydrogen bond with the Tyr 72 hydroxyl (Fig. 2H). Hence, in contrast to the PAS-bound substrates, PAS-bound choline and TCh are aligned with the dipole moment of the gorge (Fig. 2, D and H). This orientation, along with the limited inhibition of mAChE by choline ( Table 2), suggests that this choline/TCh molecule reflects the egressing product, exiting from the gorge.

Substrate Binding at Surface Sites Removed from the Gorge Entrance
At the molecular surfaces of the SCh-and ATCh-S203A complexes, strong electron densities with peaks above 4 (SCh-S203A complex) and 3.7 (ATCh-S203A complex) unambiguously account for one SCh and two ATCh molecules bound far from the gorge entry, in the region of the putative back door (Fig. 4).
In the two complexes, one substrate molecule occupies a depression delimited by helix ␣␤3,2 and the ␣␤3,2-␤2 segment in the long ⍀ loop, Cys 69 -Cys 96 , on one side, the loop connecting helix ␣(1)8,9 to helices ␣(2)8,9 and ␣(2)8,9 on the other side, and loop ␤3-␣3,4 at the edge and loop ␤8-␣(1)8,9-␣(1)8,9 at the base (Fig. 4, A and C). The choline moieties common to the bound ATCh and SCh are similarly positioned at the base of the depression, whereas the second choline moiety in SCh extends toward the top (Fig. 4, B and D). The succinate moiety of SCh adopts a more extended conformation (105°instead of 126°) than that of SCh bound within the gorge, resulting in interammonium distances of 12.8 and 11 Å, respectively. The interaction is dominated by van der Waals interactions primarily with sequence-conserved residues from the long ⍀ loop, arguing for a key role of this surface loop for AChE activity as earlier proposed from fluorescence (59 -61) and molecular modeling studies (62). Indeed, the side chains of Glu 81 , Glu 84 , Met 85 , and of the Asn 87 -Pro 88 -Asn 89 tripeptide in the long ⍀ loop, along with that of Asp 131 in the ␤3-␣3,4 loop, interact with the acetyl group in ATCh and an acyl group in SCh, whereas the side chains of Thr 436 in the ␤8-␣(1)8,9 loop, Glu 452 in helix ␣(1)8,9, and Leu 457 and Tyr 465 in loop ␣(1)8,9-␣(2)8,9 complete the binding site for the choline moiety. Several aliphatic residues, such as Pro 88 and Val 132 at the top and Leu 463 at the base, complete the shape of the cavity. It is worth noting that four conserved negatively charged residues (Glu 81 , Glu 84 , Asp 131 , and Glu 452 ) are involved in SCh binding (versus two for ATCh binding), and that the conserved pair Tyr 465 /Glu 452 , located near the quaternary ammonium group of either substrate, is reminiscent of the functional Tyr 72 /Asp 74 pair located within the PAS. The ϳ10 Å distance between this surface binding site and the conserved Asn 464 -linked carbohydrate moiety excludes direct contribution of this carbohydrate moiety for ligand binding.
In the ATCh-S203A complex, the second substrate molecule bound at the surface in the putative back door region is located farther from the gorge entrance and closer to the N-terminal region, compared with the first molecule (Fig. 4, A and B). This second molecule is positioned within a second small depression, separated from the first one by ϳ9 Å and the protruding Asp 131 . It interacts with residues that are mostly conserved, e.g. Leu 22 in strand ␤0, Leu 33 in ␤1, Phe 65 in ␤3, Arg 90 in the long ⍀ loop, and Asp 131 , Asp 134 , and Arg 136 in the ␤3-␣3,4 loop.
In both the SCh-and ATCh-S203A complexes, the surfaceexposed side chain of Glu 81 rotates, by 80°, to accommodate the bound ligand, but no significant conformational changes are detected elsewhere (Fig. 4, D and E). Should a shutter-like movement of the Trp 86 side chain be involved in opening of a back door (28), our structures do not reveal it. Moreover, the presence of bound substrate in the back door region of the S203A mutant, but not mAChE, suggests limited affinity for this region compared with the base and entrance of the active center gorge. Next, in this region bound choline/TCh is absent in mAChE, despite identical soaking procedures. This suggests that this region does not provide an exit for the reaction products. Finally, an inhibitory antibody raised against Electropho-rus electricus AChE was shown to bind to a surface site distinct from the PAS, and further identified as the region of the putative back door (63). In fact, the mAChE Asp 460 -Gln 474 segment that contains several SCh-and ATCh-interacting residues (Fig.  4), corresponds to the E. electricus AChE segment Glu 484 -Arg 498 shown, by mutagenesis analysis, to be involved in antibody binding (63). That binding of this antibody does not prevent inhibition by the substrate, as shown by the unaltered K ss value, appears consistent with the absence of bound ACh in the back door region of the S203A mutant or bound substrate in this region on mAChE.
Secondary binding sites for the PAS ligands propidium, decidium, and gallamine were not observed in structures of mAChE complexes (19) even though cocrystallization used ligand concentrations higher than those used for crystal soaking. This also argues for specific, although low affinity, binding of the substrates to the back door region.

Inhibition Kinetics and Binding of Substrates and Reversible Inhibitors
Inhibition, by SCh, 4K-TMA, and choline, of ATCh hydrolysis catalyzed by mAChE yielded constant (K i ) values close to the dissociation constant (K d ) values determined by competition ( Table 2). The K d values for interaction of SCh, BTCh, 4K-TMA, choline and of the rapidly hydrolyzed substrates ACh and ATCh with mAChE and the catalytically inactive S203A mutant, were determined from competition of the initial rate of binding of the reversible bisquaternary inhibitors, ambenonium (52) and decidium (53) (Fig. 5). These two compounds have the capacity to occupy both the active center gorge and the PAS, as shown for decamethonium or BW28C51 bound to crystalline TcAChE (34,41), although decidium may preferably occupy the PAS, as seen in the crystalline mAChE complex (19).
Ambenonium and decidium competition yielded similar K d values for most of the six ligands (Table 2). However, binding of ACh or ATCh to mAChE caused a biphasic decrease of both the ambenonium and decidium association rates, yielding K d and K d L values for each substrate (Fig. 5). The biphasic kinetics reflect formation of a ternary complex between the enzyme and associating and competing ligands. The K d value, in the midmicromolar range similar to the Michaelis constant (K m ) for ATCh, reflects interaction at the active center. The K d L value, in the millimolar range, approaches the substrate inhibition constant (K ss , determined from the dependence of the rate of ATCh hydrolysis by mAChE on ATCh concentration) and reflects binding to the PAS. Association rates of both ambenonium and decidium thus decrease upon ACh or ATCh binding to and hydrolysis at the active center and further decrease upon their binding to the PAS. These K d L values, being slightly lower than the K ss values for ACh and ATCh (Table 2), are consistent with constants derived from competition with fasciculin (11) and ambenonium (52).
A notable difference in the ambenonium and decidium competition data resides in the respective amplitudes of the two binding phases. The second phase, observed at higher ligand concentrations, is only one-fifth of the total amplitude in ambenonium competition but about half of the total amplitude in decidium competition (Fig. 5). The relative amplitudes reflect decidium association rates being less sensitive to ligand occupation at the active center than ambenonium rates, indicating that the decidium preference for the PAS more readily allows ternary complex formation. Yet, the limited specificity of decidium for PAS interactions, reflected by the similar ambenonium-and decidium-derived K d values (Table 2), and its ability to detect binding of ACh or ATCh in the active center, reflected in the first inflection point, indicate binding of its charged trimethylammonio moiety deep within the gorge, in a position suitable for direct competition with the substrate. Hence, the decidium specificity for binding to the PAS resides mainly in its higher sensitivity to PAS competition. This is particularly evident in analysis of the 4K-TMA interaction with mAChE, where the second phase of binding is detected by decidium but not ambenonium.
The similarity in the K d values for ACh, ATCh, and 4K-TMA binding to mAChE with the K m for ATCh (Table 2) is consistent with formation of a covalent bond with Ser 203 during the competition assay, as is evident during ACh and ATCh hydrolysis and demonstrated here for 4K-TMA binding to mAChE (Fig.  2B). The millisecond time interval of the competition assay for substrate interaction with mAChE allows for a few catalytic cycles of ACh (or ATCh) hydrolysis to elapse including covalent bond making and breaking.
The K d values determined for the S203A mutant, where covalent bonding is not possible, are systematically greater than those for mAChE ( Table 2). For the large substrates SCh and BTCh and for choline, the difference is only 2-3-fold, whereas for the small substrates ACh and ATCh and for 4K-TMA it is 10 -20-fold. The catalytic cycles for SCh and BTCh hydrolysis by mAChE are more than 2 orders of magnitude slower than the duration of the competition assay. Hence the K d values determined for these two substrates represent their true dissociation constants. For ACh, ATCh, and 4K-TMA, the higher K d values, being greater than the K m for ATCh and mAChE, likely reflect unfavorable binding orientations in the mutant, as observed in the ACh-and ATCh-S203A complexes (Fig. 2, C and D), and consistent with the absence of bound 4K-TMA in the crystalline mutant (see above). This suggests that the structures of the ACh-and ATCh-S203A complexes may differ from the reversible Michaelis complex of ACh with mAChE.
The low millimolar K d values observed for the 4K-TMA, ACh, and ATCh interaction with mAChE likely reflect binding to the PAS as seen in the structures (Fig. 2, B-D). In contrast, the PAS interactions of SCh and BTCh observed in the structures (Fig. 2, E-G) were not resolved in the competition binding assays. In the S203A mutant, the loss of affinity at the active site may preclude detection of active site and PAS interactions.

Implications for AChE Catalysis and Regulation
This set of eight structures, in highlighting distinctive positions, orientations, and conformations of various substrates and analogues and a product trapped in the crystalline wild type or mutant mAChE, along with the binding and inhibition data obtained on the same complexes, led us to suggest a succession of steps that may contribute to understanding catalysis and its regulation by high substrate concentrations. Complexes with an analogue and a substrate anchored to Glu 202 and His 447 within the active site pocket and oriented perpendicular to the gorge axis (TMTFA-, 4K-TMA-, and SCh-mAChE conjugates; Fig. 2, A, B, and F) clearly reflect acylation transition states and an acyl-enzyme. They also illustrate the key role of the aromatic Trp 86 and charged Glu 202 side chains in selective orientation of the ligand within the active site, through cationinteraction and longer range (4.2-4.5 Å) electrostatic interactions, respectively. Complexes of the S203A mutant with intact ACh and ATCh molecules that are rotated compared with the hemiketal conformation and still bound, via their quaternary ammonium group, to the anionic subsite of the active center (Fig. 2, C and  D), likely reflect the deacetylated enzyme structure. In turn, the choline or TCh molecule bound in the active site (Fig. 2, D, G, and H) and those bound to the PAS and aligned with the gorge dipole moment (Fig. 2, D and H), must mimic the dissociating and the egressing products, respectively.
Because of the high soaking concentrations necessary to observe PAS-bound ligands in the structures, the substrates or substrate analogues anchored, via their quaternary ammonium group, to Trp 286 in the PAS and aligned against the gorge dipole moment (cf. Fig. 2, B, C, and E-G), likely reflect binding associated with substrate inhibition. Hence, substrate in excess will slow down the exit of the choline product both by electrostatic repulsion, via its acetyl group, and by steric occlusion of the gorge entrance. In fact, these substrates or analogues may also reflect the incoming substrate, initially attracted via its acetyl group by the PAS, and now transiently occupying the gorge entrance before it proceeds toward the active center. Hence, the PAS-bound substrate would simply slide down into the gorge, to place the acyl carbon in a position suitable for attack by the Ser hydroxyl, and then reorient to stabilize the quaternary ammonium group within the anionic subsite. This motion would be associated with rearrangement of the substrate from the non-relaxed trans, trans conformation adopted at the PAS to the presumably more labile relaxed trans, gauche conformation adopted in the active center, along with major although transient rearrangements of side chains within the gorge, such as those of Tyr 337 and Tyr 124 as observed in the structures. Finally, substrate binding in the putative back door region (Fig. 4), although lacking a structurally or functionally evident portal of entry, could allosterically affect substrate binding and catalytic parameters.
In conclusion, in the crystalline mAChE and S203A mutant we identified several successive positions and orientations of the substrates bound to the PAS at the gorge entrance and proceeding within the gorge toward the active site; the conformation of the acylation transition state and acyl-enzyme; the positions and orientations of the dissociating and egressing products; and those of intact substrate molecules stably bound at secondary sites in the region of the putative back door. The lability of the substrate and intermediates in the pathway requires that surrogate structures be used in several cases. Nevertheless, these structures provide a comprehensive set of snapshots of the routes for traffic of substrates and products, the reaction intermediates along the gorge pathway, and perhaps, allosteric regulation of catalysis by substrate binding to low affinity sites at the enzyme surface.
After a revised version of this manuscript was submitted, an article was published on-line (65) that reported structural data with some common substrates and analogues and yielded similar conclusions despite differences in experimental conditions. Our study considers additional ligands, provides a kinetic analysis of ligand association, and identifies new binding sites on the enzyme surface.