Metaphosphate in the active site of fructose-1,6-bisphosphatase.

The hydrolysis of a phosphate ester can proceed through an intermediate of metaphosphate (dissociative mechanism) or through a trigonal bipryamidal transition state (associative mechanism). Model systems in solution support the dissociative pathway, whereas most enzymologists favor an associative mechanism for enzyme-catalyzed reactions. Crystals of fructose-1,6-bisphosphatase grow from an equilibrium mixture of substrates and products at near atomic resolution (1.3 A). At neutral pH, products of the reaction (orthophosphate and fructose 6-phosphate) bind to the active site in a manner consistent with an associative reaction pathway; however, in the presence of inhibitory concentrations of K+ (200 mm), or at pH 9.6, metaphosphate and water (or OH-) are in equilibrium with orthophosphate. Furthermore, one of the magnesium cations in the pH 9.6 complex resides in an alternative position, and suggests the possibility of metal cation migration as the 1-phosphoryl group of the substrate undergoes hydrolysis. To the best of our knowledge, the crystal structures reported here represent the first direct observation of metaphosphate in a condensed phase and may provide the structural basis for fundamental changes in the catalytic mechanism of fructose-1,6-bisphosphatase in response to pH and different metal cation activators.

The hydrolysis of a phosphate ester can proceed through an intermediate of metaphosphate (dissociative mechanism) or through a trigonal bipryamidal transition state (associative mechanism). Model systems in solution support the dissociative pathway, whereas most enzymologists favor an associative mechanism for enzyme-catalyzed reactions. Crystals of fructose-1,6bisphosphatase grow from an equilibrium mixture of substrates and products at near atomic resolution (1.3 Å). At neutral pH, products of the reaction (orthophosphate and fructose 6-phosphate) bind to the active site in a manner consistent with an associative reaction pathway; however, in the presence of inhibitory concentrations of K ؉ (200 mM), or at pH 9.6, metaphosphate and water (or OH ؊ ) are in equilibrium with orthophosphate. Furthermore, one of the magnesium cations in the pH 9.6 complex resides in an alternative position, and suggests the possibility of metal cation migration as the 1-phosphoryl group of the substrate undergoes hydrolysis. To the best of our knowledge, the crystal structures reported here represent the first direct observation of metaphosphate in a condensed phase and may provide the structural basis for fundamental changes in the catalytic mechanism of fructose-1,6-bisphosphatase in response to pH and different metal cation activators.
Phosphatases, mutases, kinases, and nucleases all catalyze phosphoryl transfer reactions central to biochemical processes that sustain life. The transfer of the phosphoryl group of a phosphate ester to water in most cases requires metal ions as cofactors, and proceeds either by way of a trigonal-bipyramidal transition state (associative mechanism), or through the formation of an unstable intermediate of metaphosphate (dissociative mechanism) (1Ϫ6). The metaphosphate anion (PO 3 Ϫ ) was first proposed as an intermediate in the hydrolysis of phosphate esters nearly 50 years ago (7,8). Although it exists as a stable entity in the gas phase, where it is relatively non-reactive (9), metaphosphate is unstable in aqueous solutions, and its existence is inferred only by indirect evidence (9Ϫ15). The likelihood of trapping metaphosphate in the active site of an enzyme is remote, because of the proximity of acceptor molecules in the active site. Yet an enzyme offers an advantage in that it reduces the free energy of the transition state. Hence, the active site itself could serve as a thermodynamic trap if metaphosphate, once generated, is denied access to an acceptor.
FBPase can be in either of two quaternary conformations, the R-state (catalytically active) or the T-state (inactive) (18). AMP and fructose 2,6-bisphosphate both inhibit catalysis by FBPase, the former through an allosteric mechanism (19,20) and the latter by direct ligation of the active site (21,22). Divalent metals (Mg 2ϩ , Mn 2ϩ , or Zn 2ϩ ) are essential for FBPase activity. Monovalent metals (K ϩ , Rb ϩ , Tl ϩ , or NH 3 ϩ ) further enhance reaction rates at relatively low concentrations, but can be inhibitory at high concentrations (23Ϫ25). The enzyme-mediated reaction is pH-dependent; plots of initial velocity versus Mg 2ϩ are sigmoidal (Hill coefficient of 2) at neutral pH, but hyperbolic at pH 9.6 (26). The kinetic mechanism at pH 7 with Mg 2ϩ as the cation activator is steady-state random (25), whereas at pH 9.6 the kinetic mechanism is rapid-equilibrium random (26). The catalytic mechanism is also sensitive to the type of cation activator: The Mn 2ϩ -activated enzyme uses exclusively the ␣-anomer of F16P 2 (27), but Mg 2ϩ -activated FBPase uses both ␣and ␤-anomers of F16P 2 (23). Mn 2ϩactivated FBPase, but not the Mg 2ϩ -activated enzyme, hydrolyzes the substrate analogue (Sp)-[1-18 O]fructose 1-phosphothioate 6-phosphate (28). The zinc cation is an activator of FBPase, and yet traces of Zn 2ϩ reduce catalytic rates of the Mg 2ϩ -activated enzyme (24). The above suggests alternative catalytic pathways, the dominant mechanism being determined by pH, the kind of cation-activator, and the conformation of the substrate.
FBPase crystallizes readily from an equilibrium mixture of products and substrates, and in fact the enzyme itself is active under conditions of crystallization. In past crystal structures of FBPase, only orthophosphate and F6P have appeared in the active site (29,30). Presumably, the observed complexes represent a minimum free energy under the conditions of the crystallization experiment. Reported here are crystal structures of FBPase at near atomic resolution, in which a partial reaction (essentially the second step of a dissociative pathway) has lead to the formation of metaphosphate in the active site. The crystallographic structures do not constitute irrefutable evidence of a dissociative reaction pathway, but they do dem-* This work was supported in part by National Institutes of Health Research Grant NS 10546 and National Science Foundation Grant MCB-9985565. 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 onstrate that FBPase can generate and stabilize metaphosphate at its active site. Hence, variations observed in the mechanism of catalysis by FBPase may arise from changes in the rate-limiting step of a dissociative pathway or even a change in pathway, for instance, from associative to dissociative.
Data Collection-Data from the low K ϩ complex (control complex) were collected at APS-Structural Biology Center (beam line 19BM), Argonne National Laboratory, on an SBC-CCD at 100 K, using a wavelength of 1.0 Å. Data from the high K ϩ complex were collected at the synchrotron beam line X4A, Brookhaven National Laboratory, on an ADSC Quantum4 CCD at 100 K, using a wavelength of 0.9795 Å. Data from the high pH complex were collected at APS-BioCars (beam line 14BM), Argonne National Laboratory on an ADSC Quantum4 CCD at 100 K, using a wavelength of 1.0 Å. Data from synchrotron sources were reduced and scaled by Denzo/Scalepack (32).
Structure Determination and Refinement-Crystals grown for this study are isomorphous to PDB code 1EYJ. Structure determinations were initiated by molecular replacement using calculated phases of 1EYJ, and then refined by SHELX (33). Model building and modifications employed XTALVIEW (34). Only distance restraints between covalently link atoms were applied to orthophosphate and metaphosphate, allowing structure factors to be the principal determinant of the geometry of the phosphoryl groups in each of the complexes. Fractional occupancy factors for a given atom or group of atoms were determined by refinement of thermal parameters for a series of fixed occupancy factors. Reported here are occupancy factors that resulted in refined thermal parameters comparable to those of neighboring atoms in full occupancy.

RESULTS
Three complexes (Table I) are presented here: (i) orthophosphate, F6P, Mg 2ϩ , and K ϩ (10 mM) at pH 7 (hereafter the control complex), (ii) metaphosphate, F6P, Mg 2ϩ , and K ϩ (200 mM) at pH 7 (hereafter the high K ϩ complex), and (iii) metaphosphate, F6P, and Mg 2ϩ at pH 9.6 (hereafter the high pH complex). All complexes have the dynamic loop (residues 52Ϫ72) in its engaged conformation, as determined by the location of Tyr 57 (29,30). Conditions that result in the formation of metaphosphate differ only in pH (7 versus 9.6) or the concentration of K ϩ (10 versus 200 mM). The kinetic mechanism of FBPase changes from steady-state random at pH 7 to rapidequilibrium random at pH 9.6 (26), and K ϩ is inhibitory at concentrations of 200 mM (K i of 68 mM), but activating at concentrations of 10 mM (K a of 17 mM) (25).
Control Complex (PDB: 1NUY)-Aside from its substantially higher resolution (Table I), the control complex is essentially identical to the Mg 2ϩ product complex of Choe et al. (29,30,40). Orthophosphate is clearly at the active site, and Mg 2ϩ occupies sites 1, 2, and 3 (Fig. 1A). The 1-OH group of F6P is in two positions, related by a rotation of approximately of 120°about the C1ϪC2 bond axis. In one of its positions (occupancy factor of 0.7), the 1-OH group coordinates the Mg 2ϩ at site 1, optimally positioned for an associative reaction ( Fig. 2A). Three of four oxygen atoms from P i are approximately equidistant from the 1-OH group of F6P and the distance between the phosphorus atom and the oxygen atom of the 1-OH group is 2.72 Å. In its second position (displaced conformation; occupancy factor of 0.3), the 1-OH group is turned away from in-line geometry, hydrogen bonding with only one of the oxygen atoms of P i (Fig.  2A). Mg 2ϩ at site 1 is 4Ϫ5 coordinated (the 1-OH group of F6P being responsible for the variation), whereas magnesium cations at sites 2 and 3 are both six-coordinated ( Fig. 2A).
High K ϩ Complex (PDB: 1NUX)-An increase in the concentration of K ϩ from 10 to 200 mM results in an anomalous signal at metal site 1 (Fig. 1B), indicating the displacement of the Mg 2ϩ from that site by a different atom type. The metal at site 1 has four inner-sphere ligands: One oxygen atom each from Asp 118 , Asp l21 , Glu 280 , and the phosphoryl species is ϳ2Ϫ2.5 Å from metal site 1 (Fig. 2B). The 1-OH group of F6P no longer coordinates the site-1 metal, being entirely in its displaced conformation. The anomalous signal from metal site 1 may be due to K ϩ or a trace contamination of a transition series cation (Zn 2ϩ , for instance) in the KCl. The latter is the most likely cause. The inclusion of 0.2 mM EDTA in the crystallization solution eliminates the anomalous signal at metal site 1. Furthermore, the tetrahedral coordination of the site-1 cation and the displaced conformation of the 1-OH group of F6P are characteristics of previously determined Zn 2ϩ -product complexes of FBPase (29,30). A mixture of Zn 2ϩ and Mg 2ϩ at occupancies of 0.25 and 0.75, respectively, account for the magnitude of the anomalous signal, and provide thermal parameters of ϳ24 Å 2 , equivalent to that of the site-1 Mg 2ϩ in the control structure The thermal parameter associated with the Mg 2ϩ at site 3 is higher than that of its counterpart in the control complex (30 versus 26 Å 2 ). The Mg 2ϩ cation may not fully occupy site 3. (As defined more clearly in the high pH complex below, Mg 2ϩ could occupy a site near Glu 98 at low occupancy, and not be resolved from the electron density associated with the water molecule that hydrogen bonds with Glu 98 and coordinates the Mg 2ϩ at site 3. See Fig. 2, B and C.) Further indications of weakened interactions involving Mg 2ϩ at site 3 are an increase in the coordination distance to the oxygen atom of the phosphoryl species and the concomitant decrease in the donor-acceptor distance to Arg 276 of that same oxygen atom (Fig. 2B).
The electron density associated with the ligand at the 1-phosphoryl pocket is a distorted tetrahedron. An elongated teardrop of electron density extends from a plane of electron density. Metaphosphate fits the planar density well, and a water molecule (perhaps representing a molecule of hydroxide) fits equally well to the teardrop of electron density. Thermal parameters of the metaphosphate molecule and its associated water/hydroxide molecule, both refined at full occupancy, are comparable to those of the ligating Mg 2ϩ and nearby side chains. The water molecule is 2.35 Å away from the phosphorus B, complex of Mg 2ϩ , K ϩ /Zn 2ϩ , F6P, and metaphosphate at pH 7 (high K ϩ complex). C, complex of Mg 2ϩ , F6P, and metaphosphate at pH 9.6 (high pH complex). Electron density (blue) covers metaphosphate/hydroxide anions or orthophosphate at a contour level of 3 , using a cutoff radius of 1 Å. Anomalous difference density (red) covers site M1 at a contour level of 3 , using a cutoff radius of 1 Å. MOLSCRIPT (39) and RASTER3D (35) were used for the illustration. atom of metaphosphate, being coordinated to the magnesium cations at sites 2 and 3, and is on the verge of hydrogen bonding to the side chain of Asp 74 (Fig. 2B). The electron density probably represents an equilibrium mixture of orthophosphate and PO 3 Ϫ /OH Ϫ . High pH Complex (PDB: 1NUW)-Glu 97 , which in the control and high K ϩ complexes coordinates magnesium cations at sites 2 and 3, now bridges the magnesium cations at sites 1 and 2. The Mg 2ϩ at site 1 is at least 5-coordinated and the Mg 2ϩ at site 2 remains six-coordinated (Figs. 1C and 2C). As in the high K ϩ complex, the 1-OH group of F6P is in its displaced orientation. The loss of Glu 97 as a coordinating ligand to Mg 2ϩ at site 3 is linked perhaps to the binding of magnesium cations at sites 3 and 4. A similar conformational change in Glu 97 occurs in Mg 2ϩ /Tl complexes, in which Tl ϩ at site 4 displaces Mg 2ϩ at site 3 (40). Glu 98 coordinates to Mg 2ϩ at site 4, whereas Asp 68 coordinates to Mg 2ϩ at site 3, but the binding of metal cations to sites 3 and 4 are probably mutually exclusive, as they are only 2.5 Å apart. Thermal parameters for Mg 2ϩ at sites 3 and 4 are 23 and 21 Å 2 , respectively, with fractional occupancies of 0.6 and 0.3. The electron density at the 1-phosphoryl pocket appears as an elongated teardrop extending from a plane (Fig.  1C). A molecule of metaphosphate, distorted from planarity, and a water molecule (or hydroxide anion) provide the best fit to the electron density. The oxygen atom of the water molecule is 3.09 Å from the phosphorus atom of the metaphosphate, and is within the coordination spheres of the magnesium cations at sites 2 and 4 (or 3), and within hydrogen bonding distance of Asp 74 (Fig. 2C). As in the high K ϩ structure, the electron density at the 1-phosphoryl pocket is consistent with an equilibrium mixture of orthophosphate and PO 3 Ϫ /OH Ϫ .

DISCUSSION
The control complex may represent a step on the reaction pathway, but it almost certainly does not represent the central kinetic complex (the interconversion of F16P 2 and F6P/P i at the active site). Liu and Fromm (22) determined a value of 2 for the equilibrium constant of the central kinetic complex. The electron density is consistent, however, with the presence of only F6P and orthophosphate. Conditions of crystallization and packing interactions of the crystal could perturb the equilibrium constant of the central complex in favor of products, or the crystallographic complex itself could represent a dead-end complex not on the reaction pathway. Nonetheless, the geometric relationship between P i and F6P ( Fig. 2A) is consistent with an associative reaction mechanism, the details of which have been presented elsewhere (40); Asp 74 (the pK a of which is raised by the proximity of Glu 98 ) abstracts a proton from a water molecule coordinated to the Mg 2ϩ at site 2 or site 3. The resulting Mg 2ϩ -coordinated hydroxide anion in turn abstracts the proton from a second water molecule (the attacking nucleophile) that bridges the magnesium cations at sites 2 and 3. The attacking hydroxide anion has in-line geometry with respect to the 1-phosphoryl group, and would be less than 3 Å from the P-1 atom. Data in support of catalytic roles for Asp 74 and Glu 98 come from directed mutations, which cause more than a 10,000-fold reduction in catalytic rates (31,36). The associative pathway must have a double proton transfer, because the catalytic base (Asp 74 /Glu 98 subassembly) is too far from the water molecule bridging the magnesium cations at sites 2 and 3 for a direct hydrogen bond.
The arrangement of catalytic side chains and ligands in the active site of FBPase, however, are also consistent with a dissociative pathway, and indeed FBPase can generate metaphosphate and the hydroxide anion from orthophosphate. To the best of our knowledge, the high pH and high K ϩ complexes reported here are the first direct observations of metaphosphate in a condensed phase. FBPase in its crystalline complex catalyzes the second step of a dissociative mechanism; however, the formation of F16P 2 from PO 3 Ϫ and F6P (the first step of a dissociative mechanism) may be inaccessible because of the mutual rotation of the plane of metaphosphate and the 1-OH group of F6P away from in-line geometry (Fig. 1, B and C). Evidently, the crystalline complex is dead-end with respect to the overall reaction, but clearly the active site of FBPase can at least equalize the free energies of bound meta-and orthophosphate. Variations in the kinetic and/or catalytic mechanisms due to changes in pH, chemical composition and/or conformation of the substrate, and to the type of metal activator (23-28), may stem from differences in the rate-limiting step of a disso-ciative pathway and/or a change in the type of pathway (dissociative versus associative).
The presence of Mg 2ϩ cations at mutually exclusive loci (sites 3 and 4) and the alternative ligation of cation sites by Glu 97 suggest the possibility of significant change in the active site during the course of the reaction. The Mg 2ϩ at site 4, a cationbinding site identified in Mg 2ϩ /Tl ϩ complexes of FBPase (40), may in fact play a direct role in catalysis (Fig. 3). In F16P 2 complexes of FBPase, Mg 2ϩ may appear initially at sites 1, 2, and 4. Stereoinversion of the 1-phosphoryl group may favor the migration of Mg 2ϩ from site 4 to 3. Hence, in all crystalline complexes of P i and F6P, Mg 2ϩ (or Zn 2ϩ ) is at site 3, whereas in F16P 2 complexes the preferred binding site for Mg 2ϩ may be site 4. Asp 74 then could activate a water molecule coordinated to magnesium cations at sites 2 and 4 for a nucleophilic attack on the metaphosphate anion (Fig. 3).
A change in metal cation coordination by Glu 97 could facilitate the putative migration of Mg 2ϩ from site 4 to 3 during catalysis. In a dissociative pathway, charge builds on the O-1 atom of F16P 2 , and as a consequence the Mg 2ϩ at site 1 may release Glu 97 from its crowded coordination sphere. A modest conformational change puts Glu 97 into the nascent coordination sphere of metal site 3, which in turn promotes the migration of the cation (with its attached hydroxide anion) from site 4 to 3 (Fig. 3). The migration of Mg 2ϩ between sites 3 and 4 would account for the significance of both Asp 68 (which coordinates Mg 2ϩ at site 3) and Glu 98 (which coordinates Mg 2ϩ at site 4) in catalysis. Mutations 2 of Glu 98 (36) and Asp 68 each reduce catalytic rates of FBPase by orders of magnitude under comparable conditions of assay. The dissociative pathway, as in Fig. 3, replaces the double proton transfer of the associative pathway, with the migration of a metal-bound hydroxide anion.
The loss of Mg 2ϩ cooperativity at pH 9.6 in FBPase kinetics may simply reflect a change in the rate-limiting step of a dissociative pathway. Mutations of Asp 68 eliminate Mg 2ϩ cooperativity at pH 7 2 and thereby implicate the cation in site 3 in cooperative phenomenon. The generation of the hydroxide anion at pH 7 may require the participation of Asp 74 and metal cations at sites 2 and 4 (or 3), whereas at pH 9.6 the generation of the hydroxide anion may occur with less involvement from the active site. Hence, the formation of metaphosphate may be limiting at pH 9.6, whereas the formation of hydroxide anion may be limiting at pH 7.
The hydrolysis of most phosphate esters in organic model systems occurs by a dissociative mechanism (3Ϫ6), and the same chemistry may occur in the active site of FBPase. In fact, Herschlag and Jencks (37) present a scheme (Chart II or III) that is strikingly similar to the complexes reported here in the relative placement of Mg 2ϩ , a metal-coordinated hydroxide ion and a planar intermediate. Furthermore, there are striking parallels between the active sites of FBPase and alkaline phosphatase (38), suggesting that the chemistry of model systems is broadly applicable to phosphatases. Ancestral phosphatases long ago may have commandeered the dominant reaction pathway in solution, but through evolution the free-energy landscape of the reaction coordinate may now place the associative and dissociative pathways on a near equal footing. Hence, small perturbations in the relative free energies of transition states, due to changes in metal cofactors, pH, chemical composition of the substrate, and/or conditions of crystallization may leverage profound effects on the catalytic mechanism.