Induced fit movements and metal cofactor selectivity of class II aldolases: structure of Thermus aquaticus fructose-1,6-bisphosphate aldolase.

Fructose-1,6-bisphosphate (FBP) aldolase is an essential glycolytic enzyme that reversibly cleaves its ketohexose substrate into triose phosphates. Here we report the crystal structure of a metallo-dependent or class II FBP aldolase from an extreme thermophile, Thermus aquaticus (Taq). The quaternary structure reveals a tetramer composed of two dimers related by a 2-fold axis. Taq FBP aldolase subunits exhibit two distinct conformational states corresponding to loop regions that are in either open or closed position with respect to the active site. Loop closure remodels the disposition of chelating active site histidine residues. In subunits corresponding to the open conformation, the metal cofactor, Co(2+), is sequestered in the active site, whereas for subunits in the closed conformation, the metal cation exchanges between two mutually exclusive binding loci, corresponding to a site at the active site surface and an interior site vicinal to the metal-binding site in the open conformation. Cofactor site exchange is mediated by rotations of the chelating histidine side chains that are coupled to the prior conformational change of loop closure. Sulfate anions are consistent with the location of the phosphate-binding sites of the FBP substrate and determine not only the previously unknown second phosphate-binding site but also provide a mechanism that regulates loop closure during catalysis. Modeling of FBP substrate into the active site is consistent with binding by the acyclic keto form, a minor solution species, and with the metal cofactor mediating keto bond polarization. The Taq FBP aldolase structure suggests a structural basis for different metal cofactor specificity than in Escherichia coli FBP aldolase structures, and we discuss its potential role during catalysis. Comparison with the E. coli structure also indicates a structural basis for thermostability by Taq FBP aldolase.

Aldolases are essential enzymes that catalyze carbon-carbon bond formation in living organisms. They are ubiquitous and highly abundant in pathways of intermediate cellular metabolism such as gluconeogenesis, the Calvin cycle, and glycolysis, where they reversibly cleave ketohexose sugars. In synthetic chemistry, the action of aldolases is precisely controlled by the stereochemistry of these reactions, and thus these enzymes are often used as an alternative to conventional chemical methods in biotransformations and synthetic organic chemistry (1,2) and especially in the synthesis of novel antibiotics (3,4).
Aldolases that cleave ketohexose substrates are among the most studied enzymes and, depending on their reaction mechanism, fall into two distinct groups. The class I enzymes utilize a lysine in Schiff base formation during catalysis and are mainly found in higher order organisms. Determination of the crystal structures of several class I enzymes (5)(6)(7)(8)(9)(10)(11)(12) together with biochemical studies (13)(14)(15)(16)(17)(18)(19) have provided mechanistic details for ligand recognition and catalysis in class I aldolases. Structurally, these aldolases display an (␣/␤) 8 barrel in a homotetrameric arrangement. In contrast, class II enzymes, found in yeast, bacteria, fungi, and blue-green algae, are most often homodimeric (␣/␤) 8 barrels (20,21) and require for catalysis a divalent metal cation, typically a transition metal such as Zn 2ϩ . The divalent cation functions as a Lewis acid to polarize the carbonyl bond of the incoming ketoses, thereby promoting cleavage of the adjacent carbon-carbon bond as well as proton transfer during enamine formation. Class II aldolases are activated by monovalent cations, such as NH 4 ϩ , are generally more stable than their class I counterparts, exhibit a wide range of substrate specificity, and are preferred for use in biotransformation chemistry (22,23). Their reaction mechanisms are diverse. For instance, in one class II enzyme, 2-dehydro-3-deoxy-galactarate aldolase, a phosphate anion rather than an amino acid side chain mediates proton transfer during enamine formation (24). Chiral discrimination among class II aldolases is subtle, and in the case of the stereoisomers fructose-1,6-bisphosphate and tagatose-1,6-bisphosphate, recognition and turnover depend on fine details of active site interactions made with substrate (25). Class II aldolases also represent potential targets for the development of anti-bacterial and anti-fungal drugs because they almost exclusively belong to prokaryotes, yeasts, and lower order eukaryotes.
Among class II aldolases, FBP 1 aldolase (E.C. 4.1.2) has been extensively characterized because of its important metabolic role in intermediate metabolism. The enzyme catalyzes the reversible aldol cleavage of FBP to the triose phosphates, dihydroxyacetone phosphate and D-glyceraldehyde 3-phosphate. * This work was supported by research grants from the Natural Sciences and Engineering Research Council of Canada and Canadian Institutes of Health Research (to J. S.) and in part by Cancer Center (CORE) Grant CA21765 and funds from the American Lebanese Syrian Associated Charities (to T. I.). This work was also supported in part by the United States Department of Energy, Division of Materials Sciences and Division of Chemical Sciences, under Contract DE-AC02-98CH10886. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The Substrate cleavage occurs during glycolysis, and the reverse reaction, aldol condensation, is used during gluconeogenesis or the Calvin cycle. Binding sites corresponding to both the catalytic divalent metal ion as well as the activation site of the monovalent cation were identified in the high resolution FBP aldolase crystal structure from Escherichia coli (21). More recently, the crystal structures of E. coli FBP aldolase crystal structure in complex with a triose phosphate transition state analogue suggested structural features involved in substrate recognition and processing (26). One important issue that remained unresolved was apparent induced fit movements that class II FBP aldolases undergo during the catalytic cycle and their relationship with active site binding. To address conformational changes during catalysis, the crystal structure of class II FBP aldolase was determined to 2.3 Å Bragg spacing from the extreme thermophile Thermus aquaticus, successfully crystallized in the presence of sulfate, a phosphate anion analogue. The resultant crystal structure not only defined the role of the induced conformational changes during the catalytic cycle but also provided an explanation as to the metal cofactor affinity by Taq FBP aldolase for Co 2ϩ , rather than Zn 2ϩ , which is preferred in mesophiles such as E. coli.

MATERIALS AND METHODS
Structure Determination-Native crystals of Taq FBP aldolase were grown as described (27) by vapor diffusion using sitting drops made up of 8.75 l of protein at 1.75 mg ml Ϫ1 , 1 l of 2 mM sucrose monolaurate, and 5 l of the precipitant solution containing 1.7 M ammonium sulfate, 0.1 M Tris-HCl, pH 7.5, and 10 mM CoCl 2 that were then equilibrated against 1 ml of the precipitant solution at 295 K. The crystals harvested were cryoprotected in mother liquor to which had been added 20% glycerol before freezing in liquid nitrogen. The crystals of the seleno-Lmethionine (SeMet) isoform of Taq FBP aldolase were obtained from hanging drops (27) that were made up of 5 l of protein solution containing 0.25 mg ml Ϫ1 of Taq FBP aldolase and 2 l of the precipitant solution made of 0.6 M ammonium sulfate and 20 mM citric acid, pH 4, that was equilibrated against 1 ml of the precipitant solution at 295 K. The crystals were cryoprotected in mother liquor supplemented with 15% glycerol before freezing in liquid nitrogen.
X-ray data collection for the SeMet isoform of Taq FBP aldolase and the native protein in complex with cobalt has been described previously (27) and is summarized in Table I. The structure solution strategy consisted of determining the structure of the SeMet isoform, crystallized in tetragonal space group I4 1 (a ϭ b ϭ 88.6 Å, c ϭ 164.1 Å), and then using it to solve by molecular replacement the native structure, crystallized in monoclinic space group P2 (a ϭ 99.5 Å, b ϭ 57.5 Å, c ϭ 138.6 Å, ␤ ϭ 90.25°). Multiple anomalous dispersion (MAD) data were scaled together with the CCP4 (28) program SCALEIT. Twelve of the fourteen selenium sites expected in the asymmetric unit were determined using SOLVE (29). The N-terminal SeMet was not found, probably because of positional disorder. Initial phases were calculated with the program MLPHARE (30) and were improved by 200 cycles of solvent flattening and gradual phase extension from 3.65 to 2.8 Å resolu-tion using the program DM (28). The final R factor for the phase extension at 2.8 Å resolution was 0.347. The selenium positions allowed unambiguous matching of the electron density to the sequence and construction of an atomic model by using the program O (30). The modest resolution and apparent mobility of the two loop regions (residues 134 -152 and 175-190), closing over the active site, made electron density tracing challenging.
Structure determination was substantially aided by the higher diffracting native enzyme x-ray data. The preliminary SeMet model was successfully used as a search model to determine the native Taq FBP aldolase monoclinic crystal structure by molecular replacement using the program AMoRe (28). Structure solution revealed two half-tetramers in the asymmetric unit cell of the native enzyme, and except for the loop regions, the subunits within the half-tetramers could be related by noncrystallographic 2-fold symmetry. The loops could be traced in three Taq aldolase protomers whose loop regions corresponded to an open conformation, whereas only one loop was fully traced in the remaining Taq aldolase protomer and corresponded to the loop region having a closed conformation. Electron density could not be associated with loop residues 140 -147 in the closed conformer.
Crystallographic Refinement-All Taq FBP aldolase structures were refined with the program CNS (31, 32) using standard protocols. The free R value (33) was monitored throughout the refinement. Table II lists the final parameters obtained for the native model. In the halftetramer containing both protomer conformations, electron density was weakest for residues 182-185 and 230 in the closed protomer and a R factor ϭ ¥ hkl ʈF obs ϪF calc ʈ/¥ hkl F abs  where N represents the number of equivalent reflections, and hkl represents the total number of unique Bragg reflections. b R free represents the R factor calculated for a test data set randomly selected from the observed reflections prior to refinement. The test data set contained 5% of the total observed data and was not used throughout refinement.  The temperature factors of these residues refined to ϳ40 -50 Å 2 only with the occupancy set to 0.5. The electron density was also weak for one of the two sulfate anions bound to the active site of all three protomers in their open conformation, and their occupancy was set to 0.5. Water molecules were initially identified in the F o Ϫ F c maps and screened for reasonable geometry and refined thermal factor Ͻ80 Å 2 . The tables show the overall crystallographic R factor and the free R factor for all observed reflections within the indicated resolution range. A Ramachandran plot analysis by the program PROCHECK (28) indicates that 90.4% of all residues lie in most favorable regions, and 9.6% lie in additional allowed regions. The structure analysis also showed that all stereochemical parameters are better than expected at the given resolution. A Luzzati plot indicated a 0.33 Å error in the atomic coordinates. Substrate Modeling-The acyclic conformation of FBP (atomic coordinates from Protein Data Bank entry 1FDJ) was modeled into the active site of the refined structure by superposing the phosphate residues onto the sulfate-binding sites. The substrate was oriented in the active site such that its C 1 -phosphate coincided with the fully bound sulfate anion in the closed conformation and keto oxygen oriented toward the metal cation. In this orientation, the C 6 -phosphate could be readily superimposed onto the second sulfate-binding site. 200 cycles of coordinate energy minimization were then performed at 300 K using CNS (version 1.1) without the x-ray term to relax possible bad contacts introduced by substrate modeling into the active site. The coordinates of the protein and the substrate did not deviate significantly after minimization compared with the starting coordinates.

Structure of the Taq FBP Aldolase
Protomer-The structure of the Taq FBP aldolase protomer (subunit molecular mass of 33 kDa) adopts an (␣/␤) 8 barrel fold (Fig. 1). The dimensions of the protomer are ϳ70 Å in height and 45 Å in width, with a depth of 38 Å. The barrel is closed on its N-terminal end by an ␣-helix (␣ 0 ) comprising residues 5-14. The core of the structure consists of an eight-stranded parallel ␤-strand assembly (strands labeled ␤ 1 -␤ 8 ), and each ␤-strand is accompanied by an ␣-helix. In addition to the eight times repeated ␤-strandloop-␣-helix-loop motif, the structure also contains a ␣-helix directly following a ␤-strand, in the case of ␤ 2 (labeled ␣-helix ␣ 2a ), ␤ 7 (labeled ␣-helix ␣ 7a ), and ␤ 8 (labeled ␣-helix ␣ 8a ). Furthermore, the two C-terminal ␣-helices, ␣ 8a and ␣ 8, are antiparallel to each other and create an arm from the barrel that mediates oligomerization. Within each protomer, 27 pairs of residues are involved in electrostatic interactions.
Quaternary Structure-Class II Taq FBP aldolase behaves as a homotetramer in solution with a molecular mass of 139 kDa (35) consistent with point group 222 (Fig. 2). The mutually perpendicular molecular dyads are defined as a right-handed set of axes P, Q, and R, where the P dyad is the crystallographic 2-fold axis, whereas the Q and R dyads are the local 2-fold axes. The dimensions of the tetramer are ϳ103 Å in height (along the R axis), 91 Å in width (along the P axis), and 83 Å in depth (along the Q axis). Each protomer is in contact with the other three subunits within the tetramer. The intersubunit interactions across the Q dyad are more extensive than those across the R dyad ( Fig. 2A). As a consequence, the former interface buries about three times as large a surface area (1844 Å 2 buried per subunit, representing 27% of total surface area) as the latter (582 Å 2 buried per subunit, or 9% of total surface area) upon tetramerization, whereas the interactions between the P axis-related subunits are minor (386 Å 2 per subunit, 6% of total surface; Fig. 2B). This arrangement results in a dimer of dimers within the homotetramer.
Open and Closed Conformations-Taq FBP aldolase crystallizes using ammonium sulfate as the precipitating agent at physiological pH in space group P2 or at pH 4 in space group I4 1 (27). The SeMet-substituted protein crystals are tetragonal, with two protomers in the asymmetric unit and diffraction to 2.8 Å Bragg spacing. Diffraction to 2.3 Å Bragg spacing was observed from native Taq FBP aldolase monoclinic crystals with four protomers in the asymmetric unit. The two protomers in the tetragonal cell are in the closed conformation, which contains one small loop (residues 175-190) and one large loop (residues 134 -152) that close over the active site. Cobalt in the Active Site-The catalytic cobalt in Taq FBP aldolase was identified based on its coordination and peak size in F o Ϫ F c electron density maps. During refinement, difference Fourier electron density maps were calculated from models comprising all atoms except those subsequently identified as cobalt cations. The three strongest peaks in the difference electron density maps were 24 -28 times over the noise level. These three peaks are the best candidates for the binding sites of the catalytic Co 2ϩ in the three subunits in their open conformation within the asymmetric unit. Interestingly, the protomer in the closed conformation has two mutually exclusive binding sites for cobalt metal cations (Fig. 4A). This dual binding is mediated by a conformational transition involving side chain rotations that occur following chelation by histidine residues 81, 178, and 208 (Fig. 4A). Because these two mutually exclusive Co 2ϩ cations are only partially occupied, their corresponding peaks in the difference electron density maps are only half the height (ϳ14 times the noise level) of that of the single cobalt ion having full occupancy found in the other three subunits within the asymmetric unit that display the open conformation. Ligands for the two mutually exclusive cobalt cations in the closed protomer include nitrogen atoms from His 81 (2.3 and 2.5 Å), His 178 (2.1 and 2.4 Å), and His 208 (2.2 and 2.4 Å) and one oxygen atom from a water molecule (2.3 and 2.8 Å), respectively (Fig. 4A) 2. A cartoon drawing of the FBP aldolase oligomer with point group 222. The three different molecular dyads comprise a right-handed orthogonal set of axes P, Q, and R as originally defined for the three 2-fold axes of lactate dehydrogenase (46). In A, the view is looking down the crystallographic dyad (P), while in B the orientation is looking down the molecular dyad (R). The dyads (R and Q in A and P and Q in B) are indicated by solid lines. Each protomer is shown in a different color. catalytic cation sites in all four subunits in the asymmetric unit, with peak heights ranging from 7 to 11 standard deviations above the noise level. The putative NH 4 ϩ cation sites were modeled with sodium scattering factors (Fig. 4B). The cation is coordinated by the nitrogen atoms of the imidazole ring of His 78 (with distances ranging from 2.8 to 3 Å in the four subunits) and the oxygen atoms of the side chains of Asp 80 (2.8 to 3 Å), Glu 130 (2.8 to 2.9 Å), Asn 251 (3.2 Å), and a water molecule (3 to 3.3 Å).
Novel Cation-binding Site-Taq FBP aldolase crystals were soaked in the presence of an yttrium chloride salt, which identified an additional metal-binding site (Fig. 4C). This metalbinding site has not been previously reported and is uniquely found in the protomer in its closed conformation. The metal is coordinated by two oxygen atoms of the side chain of Asp 102 (2.7 Å), one oxygen atom of the carboxylate group of Glu 132 (2.9 Å), and the hydroxyl of Ser 104 (2.8 Å), Wat 545 (2.9 Å) and contacts a sulfate anion. The metal refined to a temperature factor below 40 Å 2 . The site was modeled using a strontium scattering factor, which has a scattering factor almost identical to that of yttrium.
Sulfate Binding-Taq FBP aldolase crystallizes from high concentrations of ammonium sulfate as the precipitating agent (27). Under these concentrations, sulfate efficiently competes with phosphate or phosphate-containing compounds for binding to proteins. Two strong peaks of 17-19 times the noise level are found in the difference Fourier maps that correspond to two sulfate anions bound to the active site of the protomer in the closed conformation. One sulfate anion engages in hydrogen bonds to the hydroxyl of Ser 211 , the amides of Gly 179 , Asp 253 , and Thr 254 , as well as to two water molecules (Fig. 4A). Gly 179 resides on the small loop that closes over the active site in this protomer. The second sulfate anion is located 10 Å away and makes an electrostatic interaction with the side chain of Arg 278 as well as hydrogen bonding to the hydroxyl of Ser 49 and three water molecules.
The same two sulfate anions are also found in the three subunits in the asymmetric unit in their open conformation (Fig. 4B). However, because residue Gly 179 of the closed protomers is not available for binding to the sulfate anion, this sulfate anion does not bind as tightly as in the protomer in the closed conformation. Difference electron density for this sulfate anion was weaker (with only a half of the signal to noise ratio (ϳ10) versus to the closed subunit). The occupancy was also reduced to one-half to obtain a refined average temperature factor of 46.2 Å 2 (compared with 36.2 Å 2 in the closed subunit, where this anion has full occupancy). By contrast, the second sulfate anion in the active site binds to the open subunit in a fashion similar to that in the closed subunit (Fig. 4B).
Residues Arg 116 and His 123 at the C-terminal end of ␣-helix 4 bind a third sulfate anion in all four protomers within the asymmetric unit. The site is located at the solvent surface of each protomer and is distant from the active site. Again, the electron density was also weak for this ligand, and the occupancy was set to one-half in all subunits to obtain a refined average temperature factor of 43.9 Å 2 . In the current model, the sulfate ion accepts hydrogen bonds from the side chains of Arg 116 (2.7 Å) and His 123 (2.8 Å), as well as from two water molecules (2.7 and 2.9 Å) in each protomer. In the closed conformation, an additional sulfate ion complexes with Arg 135 side chain, one water molecule, and the yttrium ion (shown in Fig. 4C).

Molecular Details of Hyperthermostability-Electrostatic in-
teractions are the major determinant in hyperthermostability of a protein (35,36). Taq aldolase, which is stable at 90°C for several hours (37), has 27 intramolecular salt bridges in its tertiary structure. Of these, only 10 similar interactions are observed in the E. coli FBP aldolase crystal structure. In particular, an intricate network of ion pairs between ␣-helices ␣ 1 and ␣ 8 in the Taq FBP aldolase crystal structure is replaced by two single salt bridges in the E. coli structure. Glutamate residues, Glu 35 (corresponding to Glu 47 in E. coli aldolase), Glu 39 (Lys 51 in E. coli aldolase), Glu 286 (Thr 339 in E. coli aldolase), and Glu 290 (Ala 343 in E. coli aldolase) surround lysine residues Lys 289 (Ile 342 in E. coli aldolase) and Lys 293 (Glu 346 in E. coli aldolase) that results in multiple electrostatic interactions. In E. coli aldolase, residue substitution, although yielding two compensating electrostatic interactions, Lys 51 -Glu 346 and Glu 47 -Arg 335 , excludes formation of an ion-pairing network. In a sequence alignment of 18 class II FBP aldolase sequences (38), only 10 of 45 residues involved in electrostatic intramolecular interactions observed in the Taq aldolase structure are invariant.
Three intermolecular ion pairs in Taq FBP aldolase are observed between subunits related by the Q dyad. Only one of these (Arg 58 -Asp 70 ) are found in the analogous E. coli structure (Lys 71 -His 99 ). In the thermophile structure, Arg 58 is involved in a network of ion pairs involving Glu 66 and Arg 69 and the 2-fold related residues (Arg 58 , Glu 66 , Arg 69 , and Asp 70 ). Additional strong interactions were also found across subunits related by the P dyad. In particular, the side chains of Glu 7 and Arg 72 are within hydrogen bonding distance (2.8 Å) and the sulfur atoms of the 2-fold related Met1 participate in van der Waals' interactions. No electrostatic interactions were found in the Taq FBP aldolase structure across the remaining interface. Protomers related by the R dyad involve, however, a considerable number of hydrophobic contacts involving residues Tyr 85 , Leu 89 , Arg 90 , Leu 92 , Arg 93 , Phe 96 , Ala 121 , Ala 124 , Val 125 , and Val 127 .

Class II Thermus aquaticus FBP Aldolase Structure-Function
Additional structural features that contribute to thermostability are short loops and close packing as exemplified by fewer cavities (36). Pairwise superposition of the crystal structures of class II FBP aldolase from E. coli and T. aquaticus with SwissProt (39) is shown in Fig. 3; the largest root mean square deviation was 1.33 Å for 221 C ␣ atoms in common between E. coli aldolase and Taq aldolase protomer in closed conformation. The superpositions demonstrate that the Taq aldolase structure has shorter loops and a higher degree of secondary structure, creating a more compact structure that is consistent with structural attributes found in thermostable proteins. Overall, loop residues make up 158 amino acids in the crystal structure of the E. coli enzyme, whereas this amount is reduced to 125 residues in Taq aldolase. For instance, the loop connecting ␣-helices ␣ 2a and ␣ 2 in the E. coli structure consists of 10 residues (residues 70 -79), whereas only three residues (residues 56 -58) are engaged in a tight turn on the thermophilic structure. In addition, the loop following ␣-helix ␣ 3 in the E. coli structure involves 11 additional residues (residues 127-137) not present in Taq aldolase. The E. coli structure has also 10 additional residues without secondary structure at its N terminus compared with the Taq structure. By contrast, ␣-helix ␣ 7a and nine additional residues comprising the sequence 215 PELVERFRASGGEIGEAA 232 in Taq aldolase are not present in the E. coli structure. Hydrophobic amino acids from this sequence pack against ␣-helices ␣ 8a and ␣ 8 that are antiparallel to each other and create an arm that protrudes from the barrel (Fig. 1). The arm appears to mediate oligomerization because of the large number of hydrophobic amino acids making intersubunit van der Waals' interactions with the Q-related protomer ( Fig. 2A). The arrangement results in residues on ␣-helix ␣ 8a of the Q-related protomer interacting with residues residing on ␣-helix ␣ 7a . The inserted sequence of 17 residues thus not only enhances subunit stability but also promotes dimer formation and is consistent with the enhanced thermostability of Taq aldolase.
Induced Fit Movements in Class II FBP Aldolases-Class II FBP aldolases undergo induced fit movements during catalysis. C ␣ backbone traces for Taq aldolase protomers in open and closed conformation (shown in Fig. 3) suggest that the small loop (residues 175-190) and a larger loop (residues 134 -152) undergo conformational changes upon active site ligand binding. Comparison of the crystal structure of E. coli class II FBP aldolase determined to 1.6 Å (21) and 2.5 Å (24) as well as in complex with phosphoglycolohydroxamate to 2.7 Å (26), which resembles the ene-diolate transition state of the dihydroxyacetone phosphate substrate, corroborates a lid closure mechanism mediating ligand binding. In the unbound E. coli structures, these flexible loops exhibited positional disorder to various extents, and both small and large loops display an open conformation, shown in Fig. 3A. In the case of the large loop, its conformation, although open, has shifted somewhat toward a closed position compared with the more open conformation observed in the thermostable enzyme. In the bound E. coli structure, the small loop closes over the active site (shown in Fig. 3B) and adopts a conformation observed in the Taq enzyme. Furthermore, as in the Taq aldolase closed protomer conformation, a number of residues of the equivalent large loop region of the E. coli enzyme could not be traced in the ligand complex, supporting enhanced flexibility by residues in the large loop and originating most likely from fewer positional constraints as a result of the conformational change by the adjacent small loop.
What is the mechanism that is responsible for lid closure? The small loop conformations point to occupancy of the sulfate-binding site as a means by which to influence the stabil-ity of the small loop in its closed conformation and hence lid closure. The three protomers in their open conformation bind only weakly to the sulfate anion that interacted with the small loop in the closed position as indicated by partial occupancy and consistent with a high millimolar K d (1.7 M ammonium sulfate concentration used in crystallization conditions). In contrast, the protomer in its closed conformation exhibits full occupancy sulfate binding because of lid closure, indicating tighter sulfate ion binding. The protomer in the closed conformation cannot open its lid because of steric hindrance of Ile 139 with the symmetry-related Val 143 . Similarly, the lid position in the three subunits having an open conformation is also stabilized by crystal contacts. In particular, Val 143 interacts with the symmetry-related Phe 108 , and the carbonyl of Ala 144 is hydrogen-bonded to the side chain of the symmetry-related Arg164. The crystal contacts by stabilizing distinct loop conformations clearly show that lid closure enhances sulfate anion affinity. Conversely, the two distinct conformational states are not inconsistent with an induced fit mechanism whereby lid closure is promoted by ligand attachment.
The sulfate anion-binding sites have enabled the identification of the FBP binding mode in the Taq aldolase active site. The position of the C 1 -phosphate moiety in the structure of the E. coli enzyme in complex with the transition state analogue, phosphoglycolohydroxamate, coincides with that of a sulfate ion present in all protomers of our structure (Fig. 3C) and that mediates lid closure. Furthermore, mutagenesis of Arg 331 in E. coli aldolase perturbs FBP C 6 -phosphate binding (40), suggesting that the sulfate anion interacting with the equivalent Arg 278 in all Taq aldolase protomers delineates the C 6 -phosphate binding locus. Superposition of the FBP C 1 -and C 6phosphate moieties with the appropriate sulfate anion-binding sites is consistent with binding by the acyclic keto form of FBP in both open and closed conformations. The FBP orientation, shown in Fig. 4C, is free of steric conflicts, and loop closure indeed traps the C 1 -phosphate, whereas the C 6 -phosphate is able to interact with Arg 278 . Additionally, FBP C 3 and C 4 hydroxyls hydrogen bond with Asp 80 and Asp 253 , respectively. The equivalent aspartate residues in the E. coli structure, Asp 109 and Asp 288 , respectively, when mutagenized compromise FBP as well as dihydroxyacetone phosphate binding (41) validating the docking of the acyclic keto form of FBP in this orientation. The docked FBP conformation corresponding to the nascent dihydroxyacetone phosphate portion of FBP also mimics the binding observed for the transition state analogue in the active site of E. coli aldolase, shown in Fig. 3C (26). Unique to the closed conformation, FBP docking allows interaction by the C 2 keto with the Co 2ϩ cation bound furthest from Glu 132 , i.e. the solvent-accessible hence exterior binding site, consistent with a reaction mechanism where the cation is able to polarize the keto moiety. Cyclic forms of FBP did not allow superposition of phosphates moieties with the binding sites for the sulfate oxyanions.
The loop formed by residues 134 -152 has undergone, in the closed subunit conformation shown in Fig. 3B, a significant conformational change compared with the open position shown in Fig. 3A that flips it toward the active site. Leu 136 in the open subunit conformation is repositioned 9.4 Å, based on C␣ coordinates, closer to the active site and whose side chain, pointing toward the active site interior, would provoke a steric clash with the His 178 side chain, located on the small loop, were it positioned as in the open conformation in Fig. 3A. Lid closure thus requires coordinate movement of both small and large loops. The displacement by the His 178 side chain (5.2 Å using C␣ coordinates) into the closed conformation remodels the Co 2ϩ -binding site with respect to the binding site observed in protomers in the open conformation; the Co 2ϩ cation at the interior site (closest to Glu 132 ) now interacts only indirectly with Glu 132 through an intervening water molecule, Wat 545 . As a result, Glu 132 and Wat 545 are able to promote synergistic binding of the yttrium cation with the sulfate anion, implying that cations in presence of oxyanions could activate class II aldolases by enhancing the stability of the closed conformation.
Exchange by the transition metal cation between the two mutually exclusive binding sites is not sterically hindered, involving merely small side chain rotations by the chelating histidine residues. The emerging picture of the catalytic cycle in class II aldolases is therefore that of a two-stage process whereby the C 1 -phosphate stabilizes a large conformational change, as exemplified by lid closure, that remodels the active site. The active site rearrangement allows the Co 2ϩ cation to exchange between two overlapping sites (1.85 Å apart in Taq aldolase). Occupancy of the exterior site renders the Co 2ϩ cation competent to act as a Lewis acid by polarizing the keto oxygen during the catalytic cycle. It may be speculated that the conformational change and/or active site cation exchange gate catalytic events.
Metal Cation Preference-One outstanding question is the basis for metal cation preference by class II FBP aldolases. In the thermophilic enzyme, the largest activation occurs with metal cofactors Co 2ϩ and Fe 2ϩ , whereas in the mesophilic enzyme, Zn 2ϩ displays highest activation. Inspection of the active sites shows that the Co 2ϩ cation is hexa-coordinated in the open conformation in Taq aldolase, whereas at the equivalent interior site in the E. coli aldolase structure, the Zn 2ϩ cation is tetra-coordinated. Both cations coordinate the same number of nitrogen atoms from histidine residues but differ as to the number of coordinating oxygen atoms. The Co 2ϩ cation coordinates an additional water molecule and interacts in a bidentate manner with the Glu 132 carboxylate side chain rather than binding in a monodentate manner as observed in the mesophilic enzyme. The composition of the coordination sphere of two cations thus entails a greater preference for oxygen lone pairs by Co 2ϩ cation that is consistent with greater hardness of the Co 2ϩ cation as a Lewis acid compared with Zn 2ϩ (42). Residues comprising the active site and coordinating the Zn 2ϩ and Co 2ϩ cations are identical in the mesophilic and thermophilic class II aldolases, and superposition of the active sites reveals no significant structural differences. The cation binding preference thus appears to be determined by hardness of the metal cation as Lewis acid and possibly by subtle structural differences between the two enzymes. Greater active site coordination by the Co 2ϩ cation in the thermophilic enzyme compared with Zn 2ϩ suggests enhanced active site integrity at higher temperatures. A similar consideration would also apply to the reaction trajectory. The Zn 2ϩ cation in E. coli aldolase complexes the transition state analogue, phosphoglycolohydroxamate, by interaction through the C 2 and C 3 oxygens as well as the active site histidine residues (26). Stability of a similar transition state architecture in the thermophilic enzyme would be enhanced at higher temperatures by interaction with the Co 2ϩ cation. Enhanced active site integrity is supported by the ϳ15-fold reduction in activity using Zn 2ϩ as a metal cofactor in the thermophilic enzyme (35). Furthermore, the higher activation by Fe 2ϩ compared with Co 2ϩ in Taq aldolase (35) concurs with the even greater hardness of the Fe 2ϩ cation, supporting the hypothesis that hardness of the metal cofactor enhances active site integrity at high temperatures.