High resolution reaction intermediates of rabbit muscle fructose-1,6-bisphosphate aldolase: substrate cleavage and induced fit.

Crystal structures were determined to 1.8 A resolution of the glycolytic enzyme fructose-1,6-bis(phosphate) aldolase trapped in complex with its substrate and a competitive inhibitor, mannitol-1,6-bis(phosphate). The enzyme substrate complex corresponded to the postulated Schiff base intermediate and has reaction geometry consistent with incipient C3-C4 bond cleavage catalyzed Glu-187, which is adjacent by to the Schiff base forming Lys-229. Atom arrangement about the cleaved bond in the reaction intermediate mimics a pericyclic transition state occurring in nonenzymatic aldol condensations. Lys-146 hydrogen-bonds the substrate C4 hydroxyl and assists substrate cleavage by stabilizing the developing negative charge on the C4 hydroxyl during proton abstraction. Mannitol-1,6-bis(phosphate) forms a noncovalent complex in the active site whose binding geometry mimics the covalent carbinolamine precursor. Glu-187 hydrogen-bonds the C2 hydroxyl of the inhibitor in the enzyme complex, substantiating a proton transfer role by Glu-187 in catalyzing the conversion of the carbinolamine intermediate to Schiff base. Modeling of the acyclic substrate configuration into the active site shows Glu-187, in acid form, hydrogen-bonding both substrate C2 carbonyl and C4 hydroxyl, thereby aligning the substrate ketose for nucleophilic attack by Lys-229. The multifunctional role of Glu-187 epitomizes a canonical mechanistic feature conserved in Schiff base-forming aldolases catalyzing carbohydrate metabolism. Trapping of tagatose-1,6-bis(phosphate), a diastereoisomer of fructose 1,6-bis(phosphate), displayed stereospecific discrimination and reduced ketohexose binding specificity. Each ligand induces homologous conformational changes in two adjacent alpha-helical regions that promote phosphate binding in the active site.

carbon bond formation in living organisms. Their role is best known in glycolysis, where fructose 1,6-bis(phosphate) (FBP) 1 aldolases (EC 4.1.2.13) promote the reversible cleavage of FBP to triose phosphates, D-glyceraldehyde 3-phosphate and dihydroxyacetone phosphate (DHAP). The class I enzyme uses covalent catalysis, implicating a Schiff base formed between a lysine residue on the enzyme and a ketose substrate. In vertebrates, there are three tissue-specific class I aldolases (aldolase A (found in skeletal muscle and red blood cells), aldolase B (found in liver, kidney, and small intestine), and aldolase C (found in neuronal tissues and smooth muscle)), and they are distinguishable on the basis of immunological and kinetic properties (1). The catalytic mechanism has been extensively studied using class I aldolase A from rabbit muscle, and key intermediates are depicted in Scheme I.
In the forward reaction, a reactive lysine residue in the active site attacks the ketose (2) of the acyclic FBP substrate (3,4). Transient formation of a dipolar tetrahedral carbinolamine with the keto function yields a neutral carbinolamine species 1, which is then dehydrated to the protonated imine form of the trigonal Schiff base 2 (5,6). Proton abstraction of the C 4 hydroxyl initiates a rearrangement resulting in cleavage of the substrate C 3 -C 4 bond and enamine formation, shown as species 3, in the active site (7). Following D-glyceraldehyde 3-phosphate release, the enamine upon stereospecific protonation (8) forms a Schiff base and is released as DHAP by the inverse reaction sequence shown in Scheme I.
From crystallographic structure determination (9), the active site in rabbit muscle aldolase, shown in Fig. 1, contains a number of charged residues, vicinal to the Schiff base-forming Lys-229 (10), that can potentially participate in catalysis. These residues can mediate proton transfers as general acid/ base catalysts and stabilize or destabilize charges, and because of their proximity to each other, they are susceptible to electrostatic modification of their pK a values, making role assignment of active site residues exceedingly complex. Residues such as Glu-187, adjacent to Lys-229, have, on the basis of mutagenic, kinetic, and structural data, more than one mechanistic role that includes substrate cleavage, charge stabilization, and mediating proton transfers at the level of the ketimine intermediate (11). Other residues, such as Asp-33 and Lys-146, also have consequential roles in catalysis, since their catalytic activity is significantly compromised upon mutagenesis (12,13,14); the mutation Lys-146 3 Arg was shown to perturb substrate cleavage and Schiff base formation (13,15). Asp-33 and Glu-187 are within hydrogen bonding distance of Lys-146 in the native structure, and these three residues, from their spatial disposition in the active site relative to Lys-229, must make separate and distinct interactions with bound substrate. Insight into how these and other active site residues participate in catalysis and, in particular, substrate cleavage has been hampered by an absence of structures of reaction intermediates shown in Scheme I.
To examine the reaction mechanism in class I aldolases and the role of active site residues, a crystallographic study was undertaken of rabbit muscle class I FBP aldolase in complex with its substrate. Acid quenching experiments using excess rabbit muscle aldolase had previously identified a covalent complex formed at room temperature in the presence of FBP that at pH 7.5 represented ϳ50% of substrate bound at equilibrium (3). The nature of the covalent linkage was not resolved. In the same study, total FBP bound to enzyme returned to free solution ϳ9 times for each net cleavage reaction, suggesting that steps after formation of the Schiff base 2 were limiting and that reaction intermediates 1 and/or 2 are most likely the preponderant equilibrium populations. However, naturally occurring covalent intermediates with FBP have not been observed to date in crystal structure determinations of class I FBP aldolases, even in the presence of excess FBP (16,17) or by sodium borohydride reduction (18). In bacterial class I aldolases of different substrate specificities, flash freezing of recombinant native aldolase crystals to ϳ100 K trapped at acid pH carbinolamine reaction intermediates (19,20) as well as a Schiff base intermediate in a mutant enzyme form (20).
To investigate the nature of the covalent equilibrium complex in native rabbit muscle aldolase, aldolase crystals were incubated in the presence of saturating FBP concentrations in nonacidic buffer (pH 7.5), similar to those used in the kinetic studies. Flash freezing of rabbit muscle crystals briefly soaked in a saturating FBP solution trapped authentic Schiff base intermediate. Two stereoisomer analogues of FBP, (2R)-mannitol-1,6-bis(phosphate) (MBP) and (4S)-tagatose-1,6-bis(phosphate) (TBP), which form noncovalent adducts, were also trapped in the aldolase active site and provided further insight into the reaction pathway and substrate recognition. The structural analysis indicates a canonical reaction mechanism by which class I aldolases cleave a C-C bond and form Schiff base intermediates and that in the case of mammalian FBP aldolase involves a conformational change induced by active site binding.

MATERIALS AND METHODS
TBP cyclohexylammonium salt was a gift of Dr W. D. Fessner (University of Darmstadt). Prior to use, the cyclohexylammonium ion was exchanged for the sodium ion using a strong cationic exchanger and neutralized with NaOH. TBP sodium salt was used in all experiments described. Hexitol-1,6-bis(phosphate) (HBP) was prepared by NaBH 4 reduction of FBP as described previously and yields a mixture of two diastereoisomers: (2R)-MBP and (2S)-glucitol-1,6-bis(phosphate) (GBP) (21). The amount of P i in TBP using 31 P NMR was estimated at 9% and was negligible in the case of FBP and HBP. All figures in the present paper were prepared using the program PyMOL (22).
Purification and Crystallization-Plasmid pPB14 coding for rabbit muscle aldolase (12) was transformed and overexpressed in Escherichia coli strain BL21-SI (Invitrogen). Recombinant rabbit muscle aldolase was purified by a combination of anion and cation exchange chromatography and size exclusion chromatography. Aldolase concentration was determined by BCA protein assay reagent (Pierce) with bovine serum albumin serving as a standard. Enzymatic activity was monitored by spectrophotometry using a coupled assay and following NADH oxidation at 340 nm (23). Aldolase crystals were grown by vapor diffusion from a 1:1 mixture of protein solution (10 mg/ml initial protein concentration made up in 20 mM Tris-HCl, pH 7.0) and precipitant buffer (17.5% polyethylene glycol 4000 in 0.1 M Na-HEPES, pH 7.5) that was equilibrated against a reservoir of precipitant.
Data Collection and Processing-Aldolase crystals were soaked for 3 min in FBP buffer (mother liquor plus 10 mM FBP) or for 10 min in HBP buffer (mother liquor plus 1 mM HBP) or for 7 min in TBP buffer (mother liquor plus 2 mM TBP). Prior to data collection, crystals were cryoprotected by transfer through a cryobuffer solution (FBP or HBP or TBP buffer plus 20% glycerol) and immediately flash frozen in a stream of gaseous N 2 cooled to 100 K. Diffraction data were collected from single crystals at beamline X8-C of the National Synchrotron Light Source (Brookhaven National Laboratory, Upton, NY) using a Quantum 4 charge-coupled device (Area Detector Systems, Poway, CA). As control, a native data set was also collected. All data sets were processed with HKL2000 (24), and the results are summarized in Table I.
Structure Solution and Refinement-Initial phases used for model building of native and liganded structures were obtained by molecular replacement using a previously determined structure for the rabbit muscle aldolase tetramer (25) (Protein Data Bank entry 1ADO). All crystal structures belong to the monoclinic space group P2 1 and have one aldolase homotetramer in the asymmetric unit. All reflections having I/(I) Ͼ 1 were used in refinement; however, electron density maps were calculated to the resolution shown in Table I and corresponded to completeness of ϳ80% in the highest resolution shell. The structures were subjected to iterative rounds of refinement (simulated annealing and minimization) with CNS (26) and model building using O (27). Water molecules were automatically added by CNS in initial rounds and manually near the end of refinement. C-terminal regions (residues 344 -363) were traced in all subunits with the exception of residues 352-357 in three subunits and the C-terminal residue Tyr-363 in the remaining subunit that were associated with regions of weak electron density.
Ligand modeling was based on interpretation of electron density shapes of 2F o Ϫ F c and F o Ϫ F c annealed omit maps and using PRODRG for topology and parameter generation (28). Binding by FBP and HBP, as the MBP stereoisomer, were readily discernable and were associated with clearly defined electron densities in the active site. Difference electron density (F o Ϫ F c ) annealed omit maps calculated in the final round of refinement confirmed identical binding of ligands in all four subunits. Electron density associated with TBP was not clearly defined, and only phosphate groups that were visible in the active sites were refined. Final model statistics, calculated with CNS and PROCHECK (29), are shown in Table I The active site has a number of acidic and basic amino acids that can participate in proton transfers and because of their mutual close proximity may potentially perturb each other's pK a , thereby making functional assignment complex. The initial nucleophilic attack on the substrate involves Lys-229 and results in Schiff base formation with Lys-229. Site-directed mutagenesis of Asp-33, Lys-146, and Glu-187 greatly diminishes catalytic activity. The orientation corresponds to viewing roughly parallel to the subunit ␤-barrel axis and looking down into the ␤-barrel from the carboxyl ends of the ␤-strands. SCHEME 1. Key intermediates of the catalytic mechanism in class I aldolase.
codes 1ZAH, 1ZAI, 1ZAJ, and 1ZAL, respectively). The final structure models of native aldolase and covalently bound FBP, MBP, and TBP noncovalent complexes have an R cryst (R free ) of 0.167 (0.205), 0.155 (0.190), 0.167 (0.204), and 0.167 (0.210), respectively. The corresponding Luzzati atomic coordinate error was estimated at 0.18, 0.16, 0.19, and 0.19 Å, respectively. Ramachandran analysis with PROCHECK placed at least 89% of nonglycine and nonproline residues of the four structures in the most favorable region and with the remainder found in allowed and generously allowed regions, attesting to good model geometry in the structures. Errors in hydrogen bond distances, positional differences, and B-factors are reported as S.D. values and were estimated based on their value in each aldolase subunit.
Structure Comparisons-Superpositions were performed with the program PyMOL (22) using C ␣ atom coordinates of identical blocks of amino acid sequences and comparing native with liganded aldolase tetramers. Due to conformational heterogeneity among subunits in N-terminal and C-terminal regions, comparisons were performed using residues 10 -343. Root mean square (r.m.s.) deviations based on superposition of equivalent C ␣ atoms are reported in Table II. Repeated superpositions were performed using stretches of 50 amino acids to detect secondary structure elements that were conformationally invariant to binding events. Lowest r.m.s. deviation corresponded to residues 150 -250. This stretch of residues (positions 150 -250) represented secondary structure elements: ␤-strands 5 and 6 as well as ␣-helices 4 -6 of the ␤-barrel structure. Residues 158 -259 encompassing these structural elements were then used in all subsequent structure superpositions to discover regions in the liganded structures that underwent conformational changes upon ligand binding. Intersubunit variability within a tetramer was analyzed by the program Polypose (31) and yielded r.m.s. differences, based on C ␣ atom coordinates, that were less than the error in the atomic coordinates for each structure.
Comparison among structures of Schiff base intermediates in class I aldolases was made at the level of the covalent intermediate by superposing the Schiff base structure with that of the Schiff base intermediate from rabbit muscle aldolase. Superposition consisted of matching the atomic positions of Lys-229 N z and FBP carbon atoms C 1 , C 2 , and C 3 in the Schiff base structure of mammalian FBP aldolase against equivalent atoms in the target aldolase structures. In E. coli D-2-deoxyribose-5-phosphate aldolase (DERA) mutant structure (Protein Data Bank entry 1JCJ) (20), Lys-167 N z atom and DERA substrate C 1 and C 2 carbon atoms were formally equivalent to Lys-229 N z , FBP C 2 , and FBP C 3 atoms, respectively. In E. coli transaldolase B, equivalent atoms in the reduced Schiff base dihydroxyacetone analogue (Protein Data Bank entry 1UCW) (32) were Lys-132 N z and carbon atoms C 1 , C H , and C 2 of the dihydroxyacetone ligand, respectively. For the carbinolamine precursor formed with pyruvate in E. coli KDPG aldolase (Protein Data Bank entry 1EUA) (19), atoms equivalent to Lys-229 N z , FBP C 1 , and FBP C 2 were Lys-133 N z and pyruvate carbon atoms C 1 and C 2 . In the structure of the covalent complex formed by archaeal Thermoproteus tenax FBP class I aldolase with DHAP (Protein Data Bank entry 1OK4)

Conformational changes induced by ligand binding
Conformational changes were identified on basis of differences in equivalent atomic positions by comparing liganded aldolases with the native enzyme. The comparison consisted of superimposing liganded tetramers onto the native tetramer using the same selected residues in each subunit. Values tabulated are r.m.s. differences between equivalent C ␣ atoms and are given in Å.
where T is a test data set randomly selected from the observed reflections prior to refinement. The test data set was not used throughout refinement and contained 5, 7, 4, and 3% of the total unique reflections for native, FBP, MBP, and TBP, respectively. e Analyzed by PROCHECK (29).
(33), equivalent atoms were Lys-177 N z and carbon atoms C 1 , C 2 , and C 3 of the DHAP ligand. Modeling and Energy Minimization-The acyclic form of FBP was built using topology parameters from PRODRG and manually docked into the active site of rabbit muscle aldolase as a noncovalent complex. The rabbit muscle aldolase-MBP complex served as template and in the initial docking was used to align the noncovalent FBP complex. The FBP complex was then energy-minimized by 2000 steps of conjugated gradient minimization in CNS.

RESULTS AND DISCUSSION
Schiff Base-Flash freezing of rabbit muscle aldolase crystals in the presence of the substrate trapped a covalent complex in the active site under equilibrium conditions. Continuous electron density, extending beyond Lys-229 N z in each subunit, shown in Fig. 2A, indicates formation of a stable covalent adduct with FBP. The planar shape of the electron density observed about the FBP C 2 carbon indicates trigonal hybridization and is consistent with trapping of a Schiff base intermediate in each aldolase subunit. Comparison of average Bfactors between bound FBP and interacting side chains, 25.0 Ϯ 2.8 and 20.0 Ϯ 6.0 Å 2 , respectively, suggests nearly full active site occupancy by FBP. Furthermore, the conformation of the crystallized enzyme is not inconsistent with that of a catalytically active conformer. Within measurement errors, kinetic parameters of soluble rabbit muscle aldolase (not shown) were unaffected by crystallization buffer, precipitant concentration used for crystallization, or glycerol cryoprotectant, and full activity was recovered upon dissolution of the crystalline enzyme.
The Schiff base intermediate, shown in Fig. 2B, engages in numerous hydrogen bonding and electrostatic interactions with active site residues. The binding mode by the P 1 phosphate was isomorphous with that reported for the NaBH 4 reduced covalent complex with DHAP (Protein Data Bank entry 1J4E) (18) (r.m.s. deviation ϭ 0.14 Å based on equivalent C ␣ atoms) wherein Arg-303 curls around and interacts electrostatically with the oxyanion, creating a phosphate oxyanion binding pocket. In addition to the electrostatic interactions, five hydrogen bonding interactions were made with the P 1 phosphate oxyanion in the binding pocket including an unusually short hydrogen bond between Ser-271 side chain and the oxyanion (2.45 Ϯ 0.02 Å) and indicating strong active site attachment by the FBP P 1 phosphate oxyanion. The P 6 phosphate binding site includes active site residue Lys-107 whose participation in P 6 phosphate binding is corroborated from affinity labeling of Lys-107 by pyridoxal-P that abrogated FBP binding but not DHAP (35). Although both P 1 and P 6 phosphate oxyanions make two electrostatic interactions with respective residues, binding by the P 6 phosphate oxyanion is slightly weaker, since it participates in only three additional hydrogen bonding interactions.
Hydrogen bonding by C 3 and C 4 hydroxyls is very strong involving interactions with cationic or anionic active site residues (36). Notable among ketohexose interactions are hydrogen bonds made between the FBP C 4 hydroxyl and residues Lys-146 and Glu-187 shown in Fig. 2A and suggests incipient C 3 -C 4 bond cleavage. The hydrogen-bonding pattern is consistent with Glu-187 abstracting the C 4 hydroxyl proton as general base, thereby initiating cleavage of the C 3 -C 4 bond in FBP and corroborating the interpretation of enzymological data (11). Interestingly, the atoms of Lys-229 Nz, FBP C 2 , C 3 , C 4 , and O 4 as well as a Glu-187 carboxylate oxygen forms a near chair-like structure in Fig. 2A that mimics the spatial arrangement of reactants in the pericyclic transition state of nonenzymatic aldol condensations with preformed enolates (37,38,39). A salient feature is the observed noncoplanarity of the FBP C 4 atom with respect to the Schiff base demanded in the transition state with respect to atoms about the cleaved bond.
The possibility of Lys-146 acting as general base is unlikely, since aldolase is active at acid pH, and affinity labeling of Lys-146 by N-bromoacetyl-ethanolamine-P at pH 8.5 but not at pH 6.5 (40) suggests an alkaline pK a for Lys-146. Rather, Lys-146 in its ammonium form stabilizes by electrostatic interaction the resultant negative charge created on the C 4 hydroxyl ion in the transition state. Loss of the positive charge on Lys-146 would thus be critical and severely compromise activity and is consistent with acidic and neutral Lys-146 mutations that abrogate catalytic activity (13,14). Asp-33 interaction with both FBP C 3 hydroxyl and Lys-146 stabilizes FBP binding and the positive charge on Lys-146 in the active site. The negative charge on Asp-33 is reciprocally stabilized by Lys-107 at the active site periphery. Charge stabilization by Asp-33 appears to be important for aldolase catalysis, since neutral mutations of Asp-33 result in a significant loss of activity (12,14).
Carbinolamine Intermediate-The electron density in Fig.  3A shows formation of a stable noncovalent hexose-P 2 adduct in 2 B. Liotard and J. Sygusch, unpublished data. the rabbit muscle aldolase active site. The nonplanar shape of the electron density about the C 2 atom of the hexose-P 2 indicates tetrahedral hybridization and is consistent with trapping of (2R)-MBP in each aldolase subunit. Binding by the (2R)-HBP stereoisomer implies stereospecific active site recognition. The (2S)-GBP stereoisomer is discriminated against, since other binding modes were not observed in electron density maps. The quite similar B-factors for MBP and interacting active site residues, 23.1 Ϯ 4.8 and 16.1 Ϯ 5.9 Å 2 , respectively, are consistent with full active site occupancy by MBP. Different from the substrate-bound enzyme, Glu-187 forms a hydrogen bond with the MBP C 2 hydroxyl and lengthens its previous hydrogen bond to make a close contact with the C 4 hydroxyl (3.4 Å). In the MBP complex, Lys-229 N z atom is positioned perpendicular to the plane defined by MBP atoms C 1 , O 2 , C 3 and is 3.3 Å from C 2 , consistent with face si nucleophilic attack on the substrate ketose and predicts formation of the (2R)-carbinolamine intermediate obtained upon NaBH 4 reduction of the Schiff base (41). Superposition of FBP and MBP structures (r.m.s. deviation ϭ 0.13 Å for C ␣ atoms of residues 10 -343) indicates identical interactions made by the ligands with active site residues and is shown in Fig. 3B. Although binding by the P 1 phosphate oxyanion is virtually indistinguishable in both structures (0.3 Ϯ 0.1 Å) including the same strong hydrogen bond made by Ser-271 with the oxyanion (2.50 Ϯ 0.05 Å in MBP), positioning of the atoms from the C 3 hydroxyl to the P 6 phosphate oxyanion is not. The MBP C 2 carbon, because of its sp 3 hybridization geometry, expands the distance between the P 1 and P 6 phosphates to 9.6 Å from 8.9 Å in the Schiff base. As a result, the positions of equivalent atoms in MBP from the C 3 hydroxyl to the P 6 phosphate oxyanion are each shifted by 0.9 Ϯ 0.1 Å with respect to FBP in the direction of the P 6 phosphate binding locus. Active site residues in contact with FBP and MBP exhibit only slight conformational differences (r.m.s. deviation ϭ 0.3 Å, including side chains).
Binding to a rigid active site affords a structural basis for stereoisomer selectivity in rabbit muscle aldolase. Attachment by each stereoisomer to the same binding site would not be identical and would entail differences in binding affinity that are reflected by preferential binding of MBP instead of GBP. Indeed, the reduced affinity for GBP by aldolase (K i ϭ 12 M (43)) originates from a binding mode different from MBP, since isostructural binding by GBP would result in steric conflict between its C 2 hydroxyl and Glu-187 and Lys-229 side chains.
Substrate Binding-To further probe the role of Glu-187 in mediating carbinolamine formation, the acyclic form of FBP was modeled into the aldolase active site using the MBP noncovalent complex as template. To mimic incipient Schiff base formation, Lys-229 was modeled as its nucleophilic form. To maximize hydrogen-bonding potential with the attacked carbonyl, the acid form of Glu-187 was used. Slight torsional oscillation of the Glu-187 side chain in the native structure would place the carboxylate within hydrogen bonding distance of Lys-229, allowing it to accept a proton from Lys-229 and thereby making it acidic. Equally, protonation of Glu-187 can occur by a proton relay implicating the adjacent Glu-189 at the active site periphery (11).
Energy minimization of the acyclic form of FBP, docked noncovalently in the active site, is shown in Fig. 4 and is consistent with the binding modes observed for both FBP in the Schiff base and for MBP. The FBP C 2 carbonyl is oriented identically to the C 2 hydroxyl of MBP, allowing face si attack on the substrate ketose by Lys-229, shown in Fig. 4. Glu-187 in its acidic form makes hydrogen bonds to the FBP C 4 hydroxyl as  Fig. 1. B, detailed comparison of the FBP Schiff base and MBP-bound aldolase structures. Superposition was done with PyMOL using averaged subunits as described under "Materials and Methods." Yellow and cyan colors were used to depict the structures corresponding to the FBP Schiff base and MBP, respectively. The P 1 phosphate of each ligand occupies the same binding locus. Differences were observed at the C 2 atom due to its different hybridization states and from atoms of the C 3 carbon to the P 6 oxyanion. Except for a small difference in P 6 phosphate binding loci, reflected by a relative shift in the ␣-helix containing residues Ser-35 and Ser-38, positional differences by all other active site residues contacting FBP and MBP were nominal.

FIG. 4. Acyclic form of FBP docked in the active site and superposition with MBP bound structure.
The ketohexose-P 2 was docked manually by superposition onto the determined MBP structure, shown in Fig. 3A. The modeled structure was then subjected to 2000 steps of conjugated gradient minimization with CNS using topology and parameters from PRODRG. Hydrogen bonding patterns (green dashes) were conserved when compared with those in FBP and MBP enzyme adducts. The only significant difference with respect to the observed enzyme adducts is an additional hydrogen bond made by Glu-187 with FBP O 2 . The orange dash illustrates the putative nucleophilic face si attack made on FBP C 2 carbonyl by Lys-229. Orientation is similar to Fig. 1. well as to the C 2 carbonyl, not possible as conjugate base, and corroborates Glu-187 participation in proton transfers at the level of carbinolamine precursor formation as well as in aligning the substrate ketose for nucleophilic attack.
Ligand Recognition-Surprisingly, TBP was not recognized as a substrate by rabbit muscle aldolase. K i determined on the basis of competitive inhibition of aldolase by TBP (125 Ϯ 16 M) is comparatively weaker than K m for FBP (5 Ϯ 1 M) and K i of HBP (0.45 M), determined previously (42). Weaker yet similar TBP binding is reflected in the structural analysis, where partial occupancy of the P 1 and P 6 phosphate binding loci was observed at saturating TBP levels in the electron density maps (see Fig. 5). Partial occupancy of both phosphate binding loci nevertheless suggests that the enzyme can interact with the acyclic form of TBP but not apparently in a unique manner with the intervening ketohexose. Controls showed no evidence for binding by P i at 500 M in native crystals, a concentration greater than 180 M that was present in the TBP soaking solution (data not shown).
Modeling of (4S)-TBP bound in the active site using either the noncovalent or covalent FBP complex as template resulted in identical docking by TBP except for the enantiomeric C 4 hydroxyl of TBP that did not interact with any active site residue. The dissimilar binding mode observed for the TBP diastereoisomer in the crystal structure implies stereospecific discrimination by the enzyme at the level of the substrate C 4 hydroxyl. The C 4 hydroxyl makes strong hydrogen bonding interactions with Lys-146 and Glu-187 in case of FBP that are not possible for TBP if TBP was to bind isostructurally. Loss of these strong interactions thus induces a different binding mode.
TBP binding suggests a hierarchical mode of attachment by aldolase ligands, which arises from the constraints imposed by a rigid binding site. Partial binding by TBP phosphates suggests promiscuous attachment by oxyanions of bisphosphorylated structural analogues whenever the interatomic distance between the analogue phosphates matches the distance between the enzyme's P 1 and P 6 binding sites and without remaining analogue atoms introducing structural clashes. Strong oxyanion binding, particularly charged interactions made with Lys-107 and Arg-303, would assure phosphate binding independent of whether weaker interactions such as hydrogen bonding with hydroxyls have been satisfied. Such reduced ligand binding specificity rationalizes the similar micromolar inhibition constants of diverse bisphosphorylated compounds, many of which are structural analogues of FBP (43), yet on basis of their structures cannot interact identically with the active site.
Induced Fit-Active site binding induces identical conformational changes in each aldolase subunit with respect to native enzyme. Backbone displacement due to ligand interaction is localized to two adjacent ␣-helical structures (residues 34 -65 and residues 302-329) not involved in subunit contacts and comprising the wall of the active site cleft distant from subunit contacts, shown in Fig. 6. From Table II, conformational displacements of these helical regions are small yet significant when compared with residues 158 -259 of the ␤-barrel, which make up part of the active site and intersubunit contacts and results in the asymmetric narrowing of the active site cleft.
Superposition of the native and the FBP bound structures are shown in Fig. 7 and delineates a conserved network of interacting active site residues and water molecules, present in all structures, that is used to bind ligand phosphate moieties as well as C 3 , C 4 , and C 5 hydroxyls. Only positions of five residues (Ser-35, Ser-38, Arg-42, Gly-302, Arg-303), each participating in phosphate binding, are displaced by the conformational change that narrows the active site cleft. Arg-303, whose side chain projects out of the active site in the native enzyme, reorganizes and curls binding the FBP P 1 phosphate. The Arg-303 side chain conformation is stabilized by a hydrogen bond to a water molecule, shown in Fig. 7, in each subunit, which in turn interacts with the Glu-189 side chain on the opposite wall of the active site cleft. The water molecule-mediated interaction is only possible due to the backbone displacement of Arg-FIG. 5. Binding by (4S)-TBP in the rabbit muscle aldolase active site. Difference electron density shown was calculated from a 1.9-Å annealed F o Ϫ F c omit map contoured at 2.5. The green dashes illustrate hydrogen bonds. Only P 1 and P 6 phosphates were visible in electron density maps and were refined. Phosphates are present at half occupancy, with water molecules accounting for the remaining electron density, and are shown in cyan. Orientation is similar to Fig. 1.   FIG. 6. Conformational changes induced upon substrate binding. The Schiff base intermediate was superposed onto the native structure as described under "Materials and Methods." In red is shown the trace of the polypeptide backbone of the covalently bound FBP structure. Trace thickness shown corresponds to a 0.35-Å diameter. The trace of residues 10 -343 of the native polypeptide backbone was visually indistinguishable from that of the FBP-bound structure except for the ␣-helical regions containing residues 34 -65 and 302-329, which in the native structure is shown in green and blue, respectively. These regions display significant conformational changes upon substrate binding as indicated in Table II. The view is looking into the subunit ␤-barrel along its axis. Table II) with respect to the native enzyme. This water molecule is invariant and is part of three water molecules shown in Fig. 7, conserved in native and liganded structures that comprise the P 1 phosphate binding locus. The change in Arg-303 side chain conformation induces a flip in the backbone carbonyl of Gly-272 to avoid steric collision that further stabilizes P 1 phosphate attachment through hydrogen bonding with the Gly-272 backbone nitrogen, shown in Fig. 2B. The same backbone displacement further enhances P 1 phosphate attachment by promoting oxyanion hydrogen bonding with Gly-302. P 6 phosphate binding induces a slightly larger conformational change that involves the ␣-helix containing residues Ser-35 and Ser-38 (ϳ1 Å in Table II) and enabling these residues to hydrogen bond the P 6 phosphate oxyanion. Arg-42 undergoes a likewise displacement allowing its side chain to interact with a water molecule bound by both C 5 hydroxyl and P 1 phosphate. Homologous conformational changes are observed upon MBP binding, shown in Table II, indicating that substrate binding and not covalent complex formation that induces the conformational changes. Furthermore, TBP binding induces similar conformational changes. From Table II, their extents are smaller and consistent with partial occupancy of the P 1 and P 6 phosphate binding sites by TBP. The absence of significant conformational changes by all other active site residues (r.m.s. change in position of active site atoms contacting FBP or MBP ϭ 0.4 Ϯ 0.1 Å) corroborates the interpretation of a rigid active site that does not adapt to binding events except by large scale movement of secondary structures.

(ϳ0.8 Å in
Water Molecules-MBP binding and Schiff base formation both displace 14 water molecules, shown in Fig. 7. The average B-factor for the water molecules ejected from the native structure upon FBP and MBP binding was 37 Ϯ 10 Å 2 , and was not significantly different from B-factors averaged over all water molecules, 42 Ϯ 15 Å 2 . The absence of significant difference in positional disorder of ejected water molecules precludes entropic gain due to solvent displacement into the nearest solvation shell as a significant factor in promoting ligand binding. Furthermore, no water molecule was observed within a van der Waals distance of the FBP C 2 carbon in the Schiff base. The water molecule closest is 4.3 Å removed (orange dash shown in Fig. 7) and forms part of the network of water molecules conserved in all structures. This water molecule, which hydrogen-bonds Glu-189 and the P 1 oxyanion in Fig. 7, also makes a hydrogen bond to the C 2 hydroxyl in the MBP structure and suggests a candidate binding site for the water molecule expelled upon carbinolamine dehydration. The absence of a water molecule binding site within close contact of the C 2 carbon in the Schiff base is not inconsistent with the stability of the covalent FBP enzyme intermediate in solution (3). Reduction in solvent translational mobility upon flash freezing (44) could promote further accumulation of the Schiff base population, not inconsistent with the apparent full occupancy by the FBP Schiff base intermediate in aldolase crystals.
Substrate Cleavage-A multiple catalytic role for Glu-187 in FBP aldolase catalysis was advanced from enzymological data that implicated the residue in C 3 -C 4 bond cleavage as well as in proton transfers at the level of the ketimine intermediate (11). The structural analysis herein reveals Glu-187 participation in hydrogen bond formation with key reaction intermediates (7,45). A catalytic mechanism by which to promote cleavage of the C-C bond in class I FBP aldolases integrating enzymology and structure data is outlined in Scheme II.
The catalytic cycle begins by FBP binding in the active site cleft. Substrate attachment induces a conformational adjustment in two ␣-helical regions that serves to stabilize attachment of the P 1 and P 6 phosphates and is accompanied by expulsion of water molecules. Glu-187, in its acid form, hydrogen-bonds both FBP C 2 carbonyl and C 4 hydroxyl, thereby aligning the electrophilic C 2 carbon for face si attack by Lys-229, outlined in the first panel. As general acid, Glu-187 transiently stabilizes the resultant dipolar carbinolamine shown in the second panel (reaction a), and mediates stereoselective proton transfers (reaction b), that yield the neutral carbinolamine precursor 1. A second series of proton transfers catalyzed by Glu-187 as general acid dehydrates carbinolamine intermediate 1 and promotes Schiff base formation 2, reaction c. The Schiff base formed is particularly stable due to the absence of a water molecule-binding site within van der Waals distance of the FBP C 2 carbon. Proton abstraction from the C 4 hydroxyl is catalyzed by Glu-187, as conjugate base, which induces a rearrangement cleaving the C 3 -C 4 bond, reaction d. The remaining active site residues are critical for substrate alignment in the active site and afford electrostatic stabilization of charges during proton transfers. The mechanism is consistent with a trajectory of least atomic motion requiring FIG. 7. Active site changes induced by ligand binding in rabbit muscle aldolase. In red are shown the polypeptide trace and active site residues for the covalently bound FBP structure, whereas in green are shown the trace and active site residues for the native structure. The blue dashes illustrate hydrogen bonds. Water molecules in blue are those found at the same loci in native and FBP structures as well as in the MBP structure. These water molecules form hydrogen-bonding bridges between FBP and side chains of invariant amino acids: Asp-109, Arg-148, and Glu-189. In green are shown water molecules found only in native structure that are ejected during FBP binding with accompanying side chain reorganization. Water molecules shown in red are only found in the covalent bound FBP structure, one of which participates in a hydrogen bonding relay between Arg-42 and FBP. The conformational changes allow Arg-303 to curl around the FBP P 1 phosphate, whereas Ser-35 and Ser-38 participate in binding the P 6 phosphate. The orange dash emphasizes the water molecule closest to FBP C 2 (4.3 Å). These differences were also observed upon MBP binding. The orientation is similar to Fig. 1. only slight torsional librations by Glu-187 glutamate side chain.
Conserved Reaction Divergent Mechanism-The reaction chemistry underlying substrate turnover has mechanistic features that are common to diverse class I aldolases. Substrate cleaved by mammalian FBP aldolase, DERA, transaldolase B, archaeal FBP aldolase, KDPG aldolase, and TBP aldolase implicates the same C-C bond relative to the ketose moiety, and Schiff base formation necessitates a lysine residue in each enzyme. In addition, substrates recognized by each aldolase have chemical structures similar if not identical to FBP. To examine potential homology among active site residues catalyzing Schiff base formation and substrate cleavage, structures of Schiff base intermediates from bacterial and archaeal aldolases were compared by superposition with mammalian FBP aldolase. In aldolases lacking structures of Schiff base intermediates derived from substrate, the Schiff base formed in mammalian FBP aldolase was used as surrogate substrate for analysis. The structure elements common to all aldolase substrates are FBP atoms starting at C 2 and extending to the P 6 oxyanion. Shown in Fig. 8 are superpositions with DERA, transaldolase B, archaeal FBP aldolase, and KDPG aldolase.
The Schiff base intermediate of rabbit muscle aldolase when superposed onto the respective covalent complexes in bacterial and archaeal aldolases positioned the FBP molecule as a Schiff base free of steric encumbrances in the respective active sites. In the active site of archaeal FBP aldolase, slight torsional rotations (Ͻ30°) about the FBP C 5 -C 6 and C 6 carbon ester bonds alleviated a close contact. In DERA, substrate atoms from C 1 to P 5 oxyanion, including hydroxyls, coincided with equivalent atoms in FBP. In transaldolase B, KDPG aldolase, and archaeal FBP aldolase, the surrogate substrate had as in DERA its P 6 oxyanion facing out of the active site toward the solvent. This orientation is entirely consistent with substrate access to the Schiff base forming lysine from solution and warrants use of FBP both as template and as surrogate sub-SCHEME 2. A catalytic mechanism promoting cleavage of the C-C bond in class I FBP aldolases. Superposition aligned the C 4 hydroxyl of each aldolase substrate facing a possible proton acceptor, namely a glutamate, a tyrosine, or a water molecule that in turn could be activated by a glutamate or lysine residue, shown in Fig. 8. In KDPG aldolase, the putative substrate O 4 is predicted to make a hydrogen bonding interaction with Glu-45. For transaldolase B, C 4 hydroxyl hydrogen bonding occurs with a water molecule, which in turn hydrogen-bonds Glu-96 and suggests in both instances that a glutamate residue is ultimately responsible for proton abstraction, hence substrate cleavage. In archaeal FBP aldolase, superposition is consistent with Tyr-146 as the residue mediating proton abstraction at the C 4 hydroxyl, whereas in DERA, it is a water molecule activated by Lys-201.
Available data support the analysis and suggest additional participation in Schiff base formation by the catalytic entity facing the C 4 hydroxyl. In KDPG aldolase, since Glu-45 is the only residue in the active site that is capable of acting as general acid/base catalyst (19), the superposition is entirely consistent with a multifunctional catalytic role by Glu-45 that includes Schiff base formation as well as C-C bond cleavage. In DERA, the water molecule that hydrogen-bonds both its substrate C 3 hydroxyl and Lys-201 has a central role in all aspects of the reaction mechanism, including substrate cleavage and Schiff base formation (20). In transaldolase B, site-directed mutagenesis of Glu-96 perturbed cleavage activity (46), and Glu-96 was proposed on structural grounds to activate the water molecule to catalyze Schiff base formation (32). Substrate cleavage catalyzed by Asp-17, which has been proposed on the basis of mutagenesis data in transaldolase B (46), is not supported from Schiff base structures superposed in Fig. 8. A direct role in cleavage by the homologous residue, Asp-33 in mammalian aldolase, is inconsistent with the orientation of the C 4 hydroxyl in the FBP Schiff base and would require Phe-135 in KDPG aldolase to catalyze proton transfers. In archaeal FBP aldolase, a role in Schiff base formation was invoked for Tyr-146 from structural considerations (33) and by enzyme trapping of FBP as a carbinolamine intermediate (47). Furthermore, superposition of mammalian aldolase onto TBP aldolase (not shown) positioned FBP, also cleaved by TBP aldolase (30), in the TBP aldolase active site, free of steric clashes with the FBP C 4 hydroxyl within the hydrogen-bonding interaction of Glu-163. This residue is located within hydrogen bonding distance of Lys-205, which is homologous to Schiff base-forming Lys-229 in mammalian aldolase.
The structural comparison has revealed an essential constituent in the aldolase reaction mechanism. A multifunctional role by a single catalytic entity appears to have been conserved across different kingdoms, indicating significant commonality in the reaction chemistry catalyzed by each aldolase. By contrast, the nature of the catalytic entity responsible for promoting Schiff base formation and C-C bond cleavage of phosphorylated substrates by a ␤-barrel platform appears to have undergone extensive evolution, resulting in divergent solutions to similar if not identical proton transfers. General acid/base catalysis by a glutamate residue in mammalian class I FBP aldolase epitomizes a canonical mechanistic feature in Schiff base-forming aldolases in carbohydrate metabolism.